SIGE_Trabajo_de_teoría_Clasificación_de_imágenes_médicas.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "SIGE - Trabajo de teoría - Clasificación de imágenes médicas",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mUJwtvKrPWqX",
        "colab_type": "text"
      },
      "source": [
        "# Detección de COVID-19 mediante imágenes de rayos-X \n",
        "Trabajo desarrollado para la asignatura SIGE - MII - ETSIIT - 19/20  \n",
        "Óscar Jiménez Fernández  \n",
        "yoscar@correo.ugr.es"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z_9bFFBUIKog",
        "colab_type": "text"
      },
      "source": [
        "Comprobamos que tenemos acceso a la GPU de cara a acelerar el entrenamiento de los modelos."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "sXnDmXR7RDr2",
        "outputId": "30b6b29a-cfde-45aa-cb7c-fce662b8f5b2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "%tensorflow_version 2.x\n",
        "import tensorflow as tf\n",
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found GPU at: /device:GPU:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wThACVVFIR2N",
        "colab_type": "text"
      },
      "source": [
        "Damos acceso a Drive de cara a poder obtener las imágenes que tenemos localizadas en la estructura de directorios adecuada para nuestro problema, que comentaremos a continuación."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZmeU0pzMgvqQ",
        "colab_type": "code",
        "outputId": "999aa7b5-88d8-4bb7-b256-3fbaab7a1b84",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 92,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A6TFL-I1FeeS",
        "colab_type": "text"
      },
      "source": [
        "## Preparación del conjunto de datos\n",
        "En esta sección leeremos cada uno de los conjuntos de imágenes de cara a poder preprocesarlas, escalándo su tamaño, balanceando cada una de las clases, y componiendo los conjuntos de entrenamiento, validación y test, además de contruir la estructura de carpetas adecuada para el proceso de aumentación de datos desde el directorio."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ygvcw-WmF6V9",
        "colab_type": "text"
      },
      "source": [
        "### Obtención del conjunto de imágenes"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZEG0VgDcZW7t",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import cv2\n",
        "import os\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "def load_images_from_folder(folder):\n",
        "    images = []\n",
        "    filesnames = []\n",
        "    for filename in os.listdir(folder):\n",
        "        img = cv2.imread(os.path.join(folder,filename))\n",
        "        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
        "        img = cv2.resize(img, (224, 224))\n",
        "        if img is not None:\n",
        "            images.append(img)\n",
        "            filesnames.append(filename)\n",
        "    return np.array(images) / 255.0, np.array(filesnames)\n",
        "\n",
        "def write_dataset_images_to_folder(path, dataset, filesnames):\n",
        "  for idx, img in enumerate(dataset):\n",
        "    img = cv2.convertScaleAbs(img, alpha=(255.0))\n",
        "    cv2.imwrite(os.path.join(path , filesnames[idx]),img)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "stzBTAStPlxO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "covid_dataset, filesnames_covid = load_images_from_folder(\"/content/drive/My Drive/SIGE DATASETS/covid\")\n",
        "healthy_dataset, filesnames_healthy = load_images_from_folder(\"/content/drive/My Drive/SIGE DATASETS/healthy\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "77FxUx-zcXv4",
        "colab_type": "code",
        "outputId": "b89d1248-1e72-4035-cc0a-8377578f3c87",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "print(covid_dataset.shape)\n",
        "print(healthy_dataset.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(198, 224, 224, 3)\n",
            "(1341, 224, 224, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jXbKk8NfcCw1",
        "colab_type": "code",
        "outputId": "db915af2-e5ed-4612-df34-ad7279a366cb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        }
      },
      "source": [
        "plt.figure()\n",
        "plt.imshow(covid_dataset[0])\n",
        "plt.grid(False)\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQEAAAD8CAYAAAB3lxGOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOy9W4hlaXbn99/nfr+fiMiIysqsUpWqutXqbtPSNGIskBlk7BcPg2AYDcwYjxnNix4M82BZD8ZYL4OZ8aCnwRpssMEXGWxhI4TtwdAC06LpVrelqZaqKqtSeYnIuJwT536/7O2Hk7911jkVUdXdVSVlu3NDkhFx9tmX71uX//qv9a0viKJIL4+Xx8vjJ/eI/VU/wMvj5fHy+Ks9XhqBl8fL4yf8eGkEXh4vj5/w46UReHm8PH7Cj5dG4OXx8vgJP14agZfHy+Mn/PjcjEAQBP9OEATvBUHwQRAEv/F53efl8fJ4eXy6I/g86gSCIIhLel/SL0s6lfRtSb8aRdGffeY3e3m8PF4en+r4vJDAX5P0QRRFD6MoWkj6nyT9zc/pXi+Pl8fL41Mcic/puieSnrrfTyV9/baTgyD4/23ZYhAECoJAsVjM/udfIpGwv4PI4vG4JCmKIiUSCYVhqDAM7VpRFMmjN67P3/ida/jvx+Pxj5zL5+v1WpKUSCR2njMIAq3Xa8XjcS0WC7sOz8H1uCf3laQwDBVF0Ufuy/NwhGGoWCxm50v6yO/+O+v12v7u34Xn5b32n4v3y2Qy9h6r1crGOh6PKxaLabVaablcar1e2/tHUaTVavWRMQ7DUKvVSkEQKB6PKwxDG6fP64jFtr47lUrZWDE2zCHzH4vFtF6vtVgs2lEUNfev93kZgU88giD4NUm/5n6/9dybJvOTjpsE0yvHD3K9m877uOvs/x4EgdLptP3LZrPKZDIqlUqqVqv2NyYMAY2iSIvFQoeHhwrDUJPJRNPpVM1mU8vlUvP5nElVKpXSbDZTOp3WarVSqVTScrlUPB7Xer1Wt9vVeDyWJBWLRWWzWY1GIwVBoIODA11dXWk6narX62m9Xiubzapare4YqWQyqXQ6rX6/r9VqpWazqdlspul0qlgspsViodVqpWQyacrolZAx4ZmlXQMhyb47GAyUTCZNOVerlVarlRaLhWKxmMbjsZbL5Y6icc9MJqMgCJTJZDSfz7VcLhWLxXRwcKA33nhD6/VajUZDvV5PnU5Hp6enSiaT+tmf/VnVajXNZjMNh0PNZjPN53O7x2Kx0Gg0UhiGGgwGisfjKpVKymQyevr0qSaTiQ4ODtTv9/XOO+98rEzcJGO8w8edg5FKJBJKp9NKJBK6d++e1uu1Li8vNZlMdHR0pEwmo+l0qsvLS83nc52cnOj6+lp3797VO++88/im639eRuBM0l33+yvP/+Zf7Hck/Y60iwT2B+VHMQB83393Xzl/0Gv45/ICfdPn/u98lkqllMvlFI/HTZmSyaRSqZSSyaQpTCqVMk+0XC6VTqfts0KhYJ53vV4rl8tpMpkol8uZoUGB5/O5IYjFYqEwDJVMJjUajUyI4vG4ZrOZKTIoIAxDzedzSdJoNFKj0VC73VY8HrfvzudzdTodNRoNBUGg4XCo9XqtSqVixiadTmuxWJg3Xa1WWq/X5pU8ggjDUEEQaLFYaL1eq1gsStooHs/JuKzXa5VKJXvmRCJh12IMZrOZksmk3n77bd25c0er1UqNRkOtVkvr9Vrf/OY3dXZ2plwup69+9au6f/++jXmv19NoNLK5wLNjXGKxmJrNpmKxmKbTqSaTibLZrI6OjvTuu+/q0aNHHytD/vgk5feyW6vVzLtns1klk0kdHBxIkjqdjslKoVDQkydPFI/HNZlMVKvV7O8fJ/OflxH4tqQ3gyB4TRvl/zuS/u4nfekmhf84Rdv/Ln/fV9jbYOsPc/ygz8GBBwWucX8PZfkbyg4KSCQS5hmjKFI2m9VsNttRLOAfBsQLrofeIAH/veVyKWnjfTOZjH0+n8+1Wq00mUwUj8fV7/cNYWBkRqORBoOBKR4KOp1O7d6TyUSJREKpVErT6VSr1UrpdNreFeWPokipVMo8frlc1nK51Gw2k7RBDvl8XuPx2GA6aCeZTBqKQPHv3bunQqGgZrOpRqOh8/NzZTIZ/dEf/ZGur69t7o6Pj/XzP//zKhQK9r69Xs+M4Hg8ViqVMogfBIFyuZwymYyNWbfbValU0vX1tb75zW9qOBwqmUxquVwqk8lotVopk8loNBp9RDZuI+Nvks+joyMVi0VNJhMFQaBisaiDgwNdXl6a8S4Wi2q323r69KnNczabNfQURZFOT09vle3PxQhEUbQKguDXJf2fkuKS/psoir7/Sd/7JAX1Sv1x370Njv0wBuA2g3FTOMDk+1g0Ho8rlUpJ0o4HjMfjymQyFpeGYahEIqFEImGTmkwmdzwn9wjDULlcTqPRSIlEwp4hiiLNZrOd98RLr1YriyHDMNR4PNZ8Plc6nTZBR8FKpZIqlYry+byKxaKCIDDDgBJL23gdBSEkWSwWFkJI0nK5NEiOx+d/3ne9XisWiymdTu/E13g+xoZ3i8ViZvAKhYK++MUvajweK5/PKx6Pq9VqSZK+8Y1v2DPF43Hl83nFYjGdnJzo7bff1nQ61XK51HA41PX1tRaLhdLptPL5/I6xymQyyuVyqtVqSiaTGg6H+vDDD/XkyRN98MEHiqJIhUJBmUzGuBO4hCiKdpAW77NYLHbeZ98p8DMOJJFIqFwum1efTCZqNpuaTCYaDoc6Ozsz4zMcDnV0dKRYLKbJZKLxeKxarWbPcNPxuXECURT9gaQ/+CG/8xFlxjNK2zjSH14hf1BPz3WBe3z/puve9F3+9x4eL8y1gGgekRDP8RkKAjrI5/M7nlHaCkkQBKbQCBsKs69ofM53+D6IgXh3vV6rXC6r0WgolUppMBgYIYiCeuGeTCYqlUqmHFwX5AB5CNFICIQHZywYJ4wCioa3w+BJG4NYLpc1HA6Vy+V0//59C4NKpZJWq5Vms5lisZi+/e1vazAY2BwUCgVls1lFUaRcLqcvfelLqtfr5v0nk4lisZiy2azy+bySyaTJQTabNeNUr9eVSCT08OFDPXjwQKenpxaWxeNxU8DFYmGhQ7vdtmvDaWDci8WiFouFms2mjQm8ymQyMcN4eHioWq1m41mtVs3Igpi4dy6X0/X1tTmf0WhkckUYc3l5eaNM/5URgzcdXvGIn1erlV599VVlMhldX1/vwFFIttVqZcSNJK1Wqx1GXdKOFyMu9orrYfJyuTRF5gDKYzjw8vF43OJo7ue94Xq9tvjfkzooELEt1+UeXDuKIo1GIy2XS4PjsVhsh8lGYUEVKNxqtVK3292Bs0Bp4vflcmkeGJhOKJJKpYxMm8/nZuzm87kqlYoePXpkhmU0Gtl8MX+MIQYLoeQ63GexWJgCZbNZm4dKpaL79+/bOGI88a5Pnz7Vw4cPDf6mUikLneAv8vm8KpWKvvKVryiZTBo68MaKsA1klEgklMvlTMnX67X++I//WO+//75xLiCPIAg0nU4t9MlkMloul6pUKhqNRspms5pMJqrX6/YOP/MzP6Pr62utVivF43FDdoPBQMViUaVSSalUSqVSSYVCwVAehmYwGOjZs2caDodaLBaGEpDnx48fq1KpKJVK6fLyUlEUqVqt3qp3L5QR8J4cKClJ77//viSZF8lkMioWiwa7isWiyuWy7t+/ryAI9PDhQ3W7XYuPuBYT7Mk5jtlspkqlYukh4K5/pvl8vgPRfUwvbXkAUAKfeQ/IfWHTUV7OgeAqlUo7BgE47uNUYlgfZ/McPIP3zrxDIpEwZeMzmPDJZKJyuWxZCzITx8fHOjs7szGbTqemwNI2hOE5UXR+ZmwkKZPJmNGD1U8kEjo8PDSlyWQyKpfLiqJIk8nEZOJ73/uekZTj8ViFQsGMWDabVTweVy6X02q1UiKR0GuvvaZ79+4pnU5rNBppNpup3+8bH7FcLpVMJlUoFIxnAEEkEgm99957+s53vmNG05O6OAZk8+7du7q+vjY+ZL1eq9lsKh6Pq1gs6vDw0PiOXC6nVqulbrdrnh0SuVarqV6vm5yQDRoOh8b68/y8K4Y/k8koDEPNZjOVy2XF43E1Gg0zzjcdL5QR+CQoTzjgIWA6ndZyuVQul9N6vda9e/f05S9/WfF4XIPBQO12Ww8fPtTp6alZXuJR2PnlcqlCoWCCSjyMgmIY8Fb7eW4fI3tOAGgdj8eVTqdVKBR2oDDxqrRBL3gePFQul1M6nTbYh7Lj5XysDyMPskAQOJ9UI9A9n88rn8+r3+9rOp1ayJXL5VQsFu2dG42GptOpZrOZvvzlL+vi4kLlclmtVkuLxcKMBWEJ4yFtc9h+jIDeKPh8Pt9BBdPp1GLmJ0+eqN/vGxcym82Uz+dtrAlJUF4MvLRBUffu3TNjF4ah+v2+zTvjibIzT6VSSUEQ6NGjR/re976nwWCwU1sgyZAcyIf5KBQKGo/HqlQqkqRSqaR8Pm+KTfgxHo8VRZFeffVV9Xo9C89g/d944w2br36/b2OEXDDPOMBYLGbkqLQx0Pl8XqVSyYzOdDq9Va9eKCPwScc+8QXELxQK9sJPnz5VGIY6PDzUwcGBGo2G3njjDc1mM73//vt67733jL1GcKRtDInyew+P98XreGEnfvNxvySLzbznJt4lty7JIC5CTUpwMpkY4UeKjxgdhOONjOc2MJQekYAcksmk5vO5KS3P1O/3zSDN53PV63WNx2N77yiKdHFxoSAI7Jmq1aqFXhg4EA4KhQGC0GPeiH/xpu122+Ctz4snk0nl83lLmxJODYdDQ00oJN7x+PhYjUbDiLxOp6PpdGpjL22MnZ/HZDKpbDarTqejp0+fqtVqKQxDVSoVq9vw4819eaf5fK5Wq6WDgwNT2mKxqGKxqEwmo36/r8Viofl8rlQqZTAfzsLLQCKRUK/XsxoEHy7mcrkd1MV8kdU4OTnReDxWuVy2zMSjR4/UbH6kRsiOHwsjgBAzgUyYnzwOBAPvhudIpVL6xV/8RX3hC1/Qt7/9bV1cXJhg4IVQCOAVrDfEjiSD0r6AZR/2k+Lz3AOT63PjxO+pVMo8fiqVMuXzZB9Ih/NBKaT4gKk+DID4I00Hr4CSAUG9gfOcBWNcKBTs3cvlss7Pzy27QKahUCjYGHEthDORSGg6nWo8HptSeuTEz4wDhsETr/zMXJbLZStwQi4qlYpOTk7MmA8GAytoSqVSxpnwbMD2fD6v9Xqt6+trnZ6eGkRHdjgviiItl0tDDnAxy+VS1WpV9XrdoDtymEgkzKgvl0uVSiWbS7gHkOPBwYGOj49tHkGuoBVfvBRFkcbjsTkGEFGr1bJiKYqZEomErq6ubtWvF9II3JQlkLawl1RWuVw2OFar1VQul3e8LQOJgtRqNRWLRTWbTX3/+9/Xu+++q1arZYqPcDDAhBx4cw8bGXQvnCgeZCAC7A2NNwIIgjcaCLSvHyB0IJaeTCYqFoumQD7mvokPIF3oaxdg2IkV8/m8ptPpTmZjvV5bfMxnqVRKJycnGg6Hkralt8DmeDyu8XhsUBZjBZnrSc2bSpSZZz5jHFFiX1xVrVY1HA4VRZFqtZqOjo7seabTqYbD4U6RFkpEiAAHQ6rw6urK+Aieg2dlrAh/MAYYk9dee03n5+c2fnAnQRDo6urKUo/pdNrgP8gsk8kon8+r0WgoHo+r2+2q3W4riiKVSiWVy2XjJKIoUqVSMY6AMctms5ZBYTw//PBDm8ePO15II7AP9yWZNyU+TqVSKhaLNqhAZwSx0+mY8hArUpJaqVT0ta99TT/1Uz+l3//939f19bV5YZQVhZO2pJck8wjEZggTwoIi+lr0VCplCo4HlrYwNJlMmlB4iEuWAmKTmFqSxXp8j3dFqbLZrPECPB/PgyGRZN6GWBnvz3vWajUVCgULFwqFgpFTCGCn09mps6e2gHlbrVZWmMNzYBx5Njw6KMT/zBwwzxj+bDarcrmsZDJp/MloNDIITaaAd2FsMQoY7NFopHa7bedi1D3KYP69TCCjR0dHltpDsUGj9XpdpVJJnU5nJ2YHUY3HYzUaDb355puKxWJqt9tWHOXRhq+jgEshfYs8ETatVitdXV3tGNYfG2LQH0DebDarUqm0UykFZEW59kk5YuuLiwtLUVUqFeVyOUkbI0NV2K/8yq/oO9/5jj788ENjw4m7pU144XPbfB+BxpMgrHh1acsL8Lw8IxwB+V2UgDoAjAPpOOBfGIYW5sDO8zxkLnz9g0/7eeHmO3hc4OlyuVS5XFaz2TSiNBaLmVJPp1Mj2Ph/sVgol8uZVyQVuJ8x8VkVxoVQBAXDuPsUKgQez8yY53I5S4PBL1Aj4BEgqUNpW7TFMy6XS41GI1O6bDZrBhjkx/192hY0slgsLKvQ7XZNFiqViq6urpTL5TQej5XJZCxs5HpkJsrlshGYoAqIacYfIpvwDxmaTCaaz+cql8u2vkGSer2e6RBcyo+lEWAAisWijo6OVCgUlE6ndXh4aBM1GAxUKBQsXQjzjJXHg65WKw2HQ6u7Rsmn06kqlYp+7ud+ToeHh3r69KkJNlDRKz8Q1RfC4Pl8mg+v73O3noQkxge6eZjL4T3mfk0ArL9PSXoD4jMbeJJ8Pm9ogaxKNps15nq9Xtt6hOl0amWng8HAKtE8IsG4YFh9uS9Gx2c+fA0HsBxDxjyj7J7j8ZWUyACFPRgMOA8Mh3cMKBFIgjmAn6DmwKf8PJGKN2a+fcVfPB7Xa6+9ZusrMMpwLoRRk8nE5gmeptvtWmgFrwP/JG2cTyaTMQLx+vraHANywPVYv9HpdFQul80wQhKPx+OPIGt/vDBGYJ8HIKUxnU717NkzO0faCEm9XlcQBGo2m1osFqrVagZd9/PTGIRer7dTsIOnLZVKajabWq/XqtVqGgwG6nQ6plhMKkrnU4eeMGRSEGzife7JtSAJfboQ7wd09aXGxLOeICNGlmQMuvfA8BsUmEiydFUQBGYU6vW6oigyZr7b7ZpiAaebzaYhH2lLTPliKpTOQ3sf1qBIhG+sdETZUKz99G+hUDDyy3thiE6vlMgQ44PiMTdwP34dBu/kC5gwOhh2Dgwqz8CcEnIy72QHKCLie4Sr2WxWrVbLKgLx6Cgrz0y6lvEgS7FcLu07jA2LmTqdjn0XhCjpI07GHy+MEbjtQPEZHGkTD5+dnZmAt1ot9Xo9FQoFvfLKK6pWqwqCQJVKxQZsNptZjAwMxvqTtmElXDabVTqd1mQysbQOBUZAbZ+TB3ZipTm8YPj4l+W+kkzY4/G48Qveknv0gHf2kJmJ9XAVgcYo8SzNZtPy4JPJRKPRyAi04XBoi2gYN0KV2WxmqAXoK2nHM+J9fSEU9fmxWMziV8/QYxRAMNLGwIzH452l15zDOFBePBwOzWh6o8qzcV0IVMIGEA3z71OsPq2K4fO8CQuDgOa+9Bj+hfoPskw+g1QsFtXr9VStVtVoNAytwXUhE9VqVeVy2Wo7lsulBoOBZY56vZ4hFwhQnqVYLGo+nxuZ6+XgpuOFMQLe4n7c331OdbFYqN1uW8mltBGiN998U3fv3lWj0bD0mSQTvnw+b0oMlMdzkOqiVrvVan0k1y9pB7ZBmMViMfN60jZ+RLCAsr6+AEVC+eE6+Iy4FSXgHYbDoRFBHiLioXxhkicyO52OYrGYOp2OCet4PDajQ9jBO/vFUSigX8bM9VFuCmu4nick99NrhGa8E/cle8G4ENOiuCgvXtbn2H22hGcj9AO5eRnCQIBaJO1cA6LPKxP39aEK7wi5yP+QmWS1CEXhEqhiHI1GJn+lUknZbFbj8VgXFxfqdrum1L5EnbnhPalI7Pf7hozy+bxms9mPhxH4QQ8Pl/zPqVRK3W5X3/rWt/To0SN9/etf1/3795XNZs2jEI8tFgtLLTHRNObAqiMgh4eHtsqMNJsnbPxCIU8g+ZBA2ggJntXzBngwFArugMYWPkPAfSEbvRfh3WKxmBGphEHtdlvT6VTdbteUBWHyqTNfI4BBicW2a+cROAhKoDOkIEQUsN7PEUoLUsEoYgS5F6jKV3PyrP7ZOA8vzVhSZks4iFGGtISkRIGZfwwZ3/NK4+dU0s73PDnMuDKOnlwul8v2DGS1ZrOZBoOBNSopFAqqVqvKZrO6vLzUkydPzEATZnpegXflfqSN8/m8GRcajQwGg1s7Hr2QRmCfH9j/nQNlwIMCufv9vr773e9quVzq5ORExWLRBAXBQplQXMgrLDiIoFgs7pBKg8FghzFGmX0c71ejoQic5wVt/x0R8na7LUk7AovASRthLJVKCsNQxWLRDE2hUDBhBC6ORiNL4XlY78OPfc+6Px4oOPf2CkaY4tNZGEfQklcejCvj7+fAIwrgOGPIgh0E2acOQSAYHr4LVObZGCdvnP08APd5FrITnovxGR5W/nlnwrgC+ykhplS4UqkoHo9bKIZcwWeR1frggw/UarVsMREsP6EoKAsiPJFIWH2GtK21GI1GhlJvO14II7Cv4Lf9jnf1eW9iRx8ThmGo6+trvfPOO1oulzo8PNxZ2AOL7Vl1aVsGLGlHGFg+S758MBjstJ/y0BJjI21LiPfThh5uAi+99wfSQ/4hKHgylqGuVitVq1XzXJBLvV7PoD6CGovFlMvllM/nDZISAlGIgodDAXx60RsjIDLXBaZL27QU8N5XYzLmIDfe0YcLhC0emjNfHlV5pMF5/v50TmIuuZ4fd7yqRxHeEPlUIOPin4t/Pt1KxeNkMjGlnc1mtkISCE82QJItVSZcG4/HGgwG5hgYf9AR9TKkStfrtXFZGP5CoWDVp5RYc7/940c2AkEQ3JX030k6lBRJ+p0oin47CIL/TNI/lNR6fupvRpveArceiURClUrF+rvdxg9I29TZfpWdj78xEsPhUA8ePFCv19Pdu3dtxdU+nEMpId2iKFKv17OyUSwtSsTS2ul0qtFotMNOe0Tg0cn+s/HMkI1wF7D8eGZSWAhfs9nUfD5XtVrVYDBQGG4ahSwWC2O9vWdFIfx9eRfSbD6d6YlOahkYG+AuYyJt6yi4VxiG9v9+5aaP3Vl3z8IsjAi/cz3QSa/Xs2dmjP2cg9bgDkBrFCl5zkH6KOHs58NnAeAsIPw4H6gvyQhbDIskK+/u9XrWr8GvM0gkElboRMUi1yQM9O+6/zN6w/oIruNrIpCtTzo+DRJYSfrHURR9NwiCoqQ/DoLgXz3/7J9HUfRPf9ALJZNJvfbaawrD0JpAMhCSdlI5CC6K6T0WAuEJm9lsplarpclkYoYAT+LzyuRyMQ4+z+4Xj0i7MR4pm+FwaI04ERQUDyH0HAaEIs+KIEImSVumP5lM7qxIS6VSurq60nK5VLfbNW+O4fAKj2Lgqebz+Q7RxXuhlJSe+hw5yoRC+MIjT+r5Ipx8Pr8Td6OUEKMYXs97SLJyZJ7d8wUYceSBvok+/cr3uTdjQk0DvQT20aVHgzwrCstnKB8H6V7SdKA5vj8cDtVoNKx60yuxr+/o9/t2TT82vi6Azxh7ZHI0Gllo0G63VavVNB6PDdnAcX3c8SMbgSiKziWdP/95GATBn2vTavyHPuLxuF555RUr7YXEAK4FQbCzQAWFwjAQ73qYjUVk0Hq9nmazmY6Pj3V8fGzxNl6La3JfT9hFUWRcAQJB5V6lUrGqOQo7SC0Cdb2BkrYNU0AydP8F0kuy+zUaDTMYYRjq6upqRwg4H0Hw+XxCJdJXvKev7MPjgJzi8bi9JwrIHOynm8jhE6J4I+zbkUXRpqmF97Sj0cjCFw/NYclBZHhYjBptwJhj0AkNPFarlWV/vKxIMmRDGhNOYz8c5Lm9Z/c1GKBPsiU+XUuhWhiGOjk5sfHDSKxWK1vht16vLd3qOQgfZjK2vvYC5JPNZq0PZBBsaj9IYZLm9e912/GZcAJBENyX9G9I+pakvy7p14Mg+PuSvqMNWuh+3PfxRFEUqdlsWqoDowBDyjp3BsGn16SNF2m32wYVvTdOpVJaLBa6uLjQfD7XvXv3TBlRQgglz7xjhfk+937+3paLBeL63gPEZwiJzygQD/Nu5KElqVwuW7yHkrTbbRMwz7hzT96Tief9iUUxHJ64I4728TAGjzw/yonycR5jn0wmrUcB92BO9sM2SVamC2SHyYdNR7kZe56t3++rWCxaERNhDqgxl8vp8PDQFAlkQ8cp0IpHV4QQHMwRXlrapgt9ahMjQZztQxkIzFdeecUQJtWhyM7V1ZWurq7M2DIPOBB4Iow/aIZnZM5zuZxOTk4s84M8YyxxDJ972XAQBAVJ/4uk/yiKokEQBP9C0m9pwxP8lqR/Jukf3PA923cgk8no6urKmO5kMqmTkxOVSiXrokpJpa/+CoJgp5qqXC6rWCxaL30GgQnkGI/Hevz4se7du2fkGAQQ95C2DKtPH1FN58kyiDnuUyqVzEuTywfC4jmB5lSRkUOnOAZoyzp7mm0gDJIsDQj09uktUIgvWGIlGzAUYzUejy3MgAj02Q1P3CG4KKk3uCgZy3N9+owUK6lGwgdCAV/0RLEQY+S9IHlvUIQPWc7OzpRMJq0vH4iCufSwmnAB9ODDON7dc03PZXZH5jCIFJQ1m00lk0nbgwBFpR6iUCjo6dOnRtryjIQdIAj4A+bEhyc+dBwMBiYryAKZHBZ4+dT4bR2HP5URCIIgqY0B+O+jKPpfJSmKokv3+b+U9Ps3fTdy+w6USqUIC0/r6lqtpuPjY1P4y8tLs5RASuAgAkmZbKVS2bHaQH4mkbLOhw8f2vptSSawXIdJxLv6arh9lhkjQaEISgmRSasnn6pi5Rp14qVSyQzDe++9p3w+b0typU0IcnR0pF6vZww/3hEDBZlJnOu9LFDZGwfQBx7JZyw8fEcJUB4QBkw4go4nxyCSreC9MIgoAkKOQeD/yWSiQqFg9QhwMMDo4XBoC7CkbV9IQh16DsTjcVuXwLUh/lBwuCAPmZkjiFBpGx6geJwDSqrVaoYY2QSkUCioUqloPp/r4uLC6vx9eTslwDxfLLbtmMVc4STIymAoyFSBknlP5gKk+7mEA8EGLxQ4T68AACAASURBVP3Xkv48iqL/0v39znO+QJL+lqR3fpDrIRisVOt0Ojo8PLT86P3793V+fm6QF5KHlwPyUetOmhCridIipMTPT58+VaPR2Okrz3l4CH8vr9w+5gX+wnT7ohgfQ4I6YrGYyuWyrYYLw1DPnj3T06dPzWMBW4mTfSOVUqmk8/NzrVYr2wiE90dJQQYeGkvbDAvviqABw3kHDCbekkKUMAyN7GLcKLP2hCyZC6A7gs5z8Z4siPIHUHq9Xpui+12P6KEXRdu1FXhov6ISJLO/NNzDZQwdPAlo0FdMenIah+KVrVwuG48DQmThGwuxeH7KxnFKvp7ChwiESlRC4oAgK+v1uo2ztG3GCsrACPlnv+n4NEjgr0v6e5L+dRAE/+/zv/2mpF8NguCr2oQDjyT9o0+6UCy2qXfGmtPUgRZPWLhKpaJ0Om3FL9SmI8woI11fWRpKfO8V0sPm6+tr9Xo9HR4eGpHo43gP9zFKeBPP+PtwhQnx3sLHluPx2BY/tVottdttyzRIMqi6XG661rDgB+a32+2qXq8rl8uZJwD5AFF9cQmK7WN0lIbuNr5foKQdKOyZcE8IIsB4MAQyCALrIERc7usMgN30QyCOBdXgqXkfuv3gLRkbSM/9smFpWx8wGAzMMPqGKZ4rwPD71J+07RsgyVKtGCUIPIrRpA3nQSrZL/zCAXzpS18yh7dcLm3VKrLhnYgky2ZJMpjPtQeDgWq1mm0ww7slk0lzBl6Objs+TXbg/5F0U0L/h9prQNpCLwpZ8Ez9fl/Hx8d69913FY/Hd+r8UUa8L0qOcJELRwDxUCgu8Sswa7lc6unTp3rttdeUz+fN2nJNPB/15/tpwP1iDjwxnoXJJV5mJdmHH35ocTIpH4T08PDQWPR9plrarrQkDUQKDQ/u+/BB8iUSCSujxvtTj1AoFCyWrFQqO6ktDEoUbRqNgMp4Dl/CS0gVi202wMAQodg+dYphQMH4nXGgzTqKzVhjdAkrGGufzQAFkjnAo5Oj39+CDU4G5yLJxtOTncgY9f/E4oQ3fjMSiFYUNJFIqNPpWLxPRyb/bJ48JtOETFMu7MM41hxQRARHQc1JPB7/8Wg0enJyYnEQsT8LK1jei3cGokHSeQ+EdUfAfTERMSOw1ysy3un09FTValWlUmknLQVvgKLjAWGGgYXkZIF7GAEP+3u9nh4+fGjNOTxr7Tv2dLtdrVabZh4U5/hFM8BDjN1kMrGVaf7zdDptZCbexldagsDw9iAg3t8bWIyub4YqyeoPfMWlb4aCV2SlG0bAr83HWGD0fOoPpeA7sPIoDaQl6MbzBolEQu1229p3EaZQCUqbLpQbJ+NX4UmyzANoAMMCmorHN+29KdVmDuGSRqORyc14PFaxWFS9XjdHxTwTMmJYkHueEeOCocFwepSLYUAWQS83HS+EEQiCwBY/UBPv+9rD4POC1EdjiSeTiVarlcVHCAECiaACyXxGAWHEg4M4Li4udOfOHRMwrDnP4glBkIwnDCmCYU24tFGUs7MznZ+fm6BjrPwCF+/58erEi8BnJhiBxsuwmhJoD5yFHPLwHcPgK9mIc1Ey0BCKS8qLnoQ+XPI8CWlD5tIXcEnbdQSgDcI4iDsKsA4ODnY8ts/UeEIWxwAKYww9SpzNZhZ24mFrtZr1KvTpVc8j8X6SzOhgvEEIicSm/v/6+lr9ft/Ye8Yfg7hYLHYyB5J0//59nZ2d7WRdkCd4GEm2NgAD4NEZ5DKIjucmJPIt0/eP21cV/CUfzWZTd+7cMU8Uj8eNyR6NRkYCsVUWngtBkGQeg5f3VhKo5gkgaZsb9kw51768vLRy1VqttlNCSh04k4xXReghizAcp6entiMu95jNZraGH27Dx6jAWQ7vhVAiFIbnp+accAnhwChIW2OF4ku7TTVBU6RMeUfYeLwz3jqZTNpCK2kjrEB6PB5zlkqlbEktc4chkmRpQeoCJO0Qgxh+3pEQIIq2zWHZo49n5D3hHDC80+lUV1dX6vf7O6lP34nYoyB6GjKO7MeAQsLtgC78OdJ2XQNzfXFxYcgHcpQ5JwuBg4C/ovGrz1gFQWDNSZlHFophQHxqef94IZDAcrnUhx9+qDt37thmmAjXxcWFpciazabq9bqePXtmxoLYx+dyQQa+8AOE4GGUT/ftk0rULMznc11dXenk5MQ8K/f1hS0Io09JxWIxPXv2TKenp4YYPEHpKwqx6nTeIbYFfrOpBUgliiLL+6MsLI8mm0HzDbwJVYGkCj2D7g3Jcrm05iP7RonKQDwb4QaEYa/X+4j3BPKTUqTmAbS0Wq2ssxFKAsxerVa209BqtdKTJ0/MEKAAnusg7KAQy6cBKTqiuAjvCJKiBsE7DcaAYh+QGHF7tVrVdDpVq9UyghujkkwmzeunUimryATF0UWYMIRwh7H2JCv8CwbBcxyZTMYyBfto0e/NeNvxQhiBMAzVarV0fn6uO3fumLAz+IeHhxqPx9a08c6dO1Za7AkYPBgCTvzIPzyOj3mB8nhYlIvP8Sbk5ulmRJw2Go0stehz1p1Ox5pBeIYdQ4AlBxGw6gzm2DPS5XLZ0p1+7QRMNPEoaS34AwjQMAx3Vj4uFgsbo8ViYSEYnhpkAvHqi6kQJhq1SrIaDcaSqkvqM7w3hsnn/WH3QTG8OzX36/VaJycnWiwWur6+Ns9IUxXGnPm+vr5WpVIxg0SY4PkfX6Dl1/5zHrCdjBRkG0ZPkpUmP378WNK2zoK9EPDizWbTqvlYFgzXAFnsq2B5P8aSUMivGvQp1XK5bKXnXg+ouQEB+lZw+8cLYQSk7WKZx48fKwxD1Wo1SZuUCzvfsBMOzRSJ41nR50nBdDqtg4MDnZ2d7aSAgHwIN/8orOBzBg3IySKd6XRq21QD3yDZMArdbtc2nMSa+1ywr5YjdPAchTcQ+Xxef/EXf2HpURp/cq0oinR1dWWNUOBMisWiCS0pU+JDkAtjRnWaJ758KSwGETK2Wq1aS3dCEpSLuSR+pb32Ps9DbM098GIgJcaX2ojlcqlarWZCDQnGeShArVaz+/uajEQiYYaL3D1OAl4iFovZmn48MwVYsVjMOBQqWT23BDqiBgPF5b14z/PzcxsPitI8j1IsFo3g9MgQLiuKop09M5HTbDZrPSRABYwzsnnb8UIYAbwEjGu/37dYhvTQdDpVv983z3dwcLAT4xL/I0Qs6qlWqwafgVR4ff99iDe8oYeZKC6WutvtqlgsKpvNqtlsWmELteyXl5cfYdP9mnSeYzAYmOflHTFwvqQUrgGBmEwm1j2Ge6AIHspjJCXtZEW8p/MLqfgOKMcvZYZHqdfrVq5Kyy5WHXIO4YxfNOO9Ec9JAc1isVC1WrXScMKbINg0iAFpAXPX67Xu3r1rRoXqQpAQfADem1QlBylJuCNkYLVa2Xk+jADaEy7B8t+WTeCeNAUBhQZBoMPDQ3sODKtfn8J4g9aQbZzTYrFQr9fbIbOLxeIOYgHpeGSBUb/peGGMAKk1Nrd49913dffuXUmylXDsJrtarXR2drajhLTF9ooOQdVoNNTpdEzQmRBpu2kmMSICQszu2Wdi7XQ6bXu7YaGvrq40Ho/NaMG0+1JPjIC30IPBwOA0UNh3BiIEoZaAeLXT6ZgQYUB8nMwzc77nPGD392Eyc0DhCYYCtAAfQa4bPsBnWWDjCX0wzKArv5QZw0WmgT0fuB9ZA4jC8Xhs+x9+8MEHunPnjmq1mvL5vC0kKpfLyufz1g9C2tah+MwCZdcYl9FotFMqnc/nlUpt+gEwp9Q8cA6hHhwHRgOkiQFEgX16Dzn1HM9+ug9Og3nEkPqsyGAwUK/X01tvvaWjoyMbRxwFCMovlNo/XojsQCqVMk8qbQt6Wq2W4vG47SnnrRnCORwOjdghx8o2ZT41c3x8vLOc1cdgTCjQHEYdD5tKpQyGHhwc2AKV4XCo999/X48fPzbCCU+FIPvVeqADDBAQGVKN9BWhTzy+6fRbrVZtPYSv8iuVSlbl6DeqJBYErSDcni1m3QEZBPgQ4LFHMowPNQtAXF8RR6jhSUPq5IG8vr4fT8fzQCIeHBzYklg85P3795XL5WyLbZRlMBjo4uLCwikMOkgA5yDJwj6Mlq/MhICEMK1Wq1qv12q1Wma8MOAeJXEtZAbUB88jyRAKBggjSWgLaUg45+WYOgiMBgZuf6FZFEV68OCBGftqtWq1LqCpjzteCCSA0lDlV6lUNBgMjHQjHYeX7Ha7qtVqBtPj8bj6/b4ZgNVqpU6nY7XnCB4NOIGYwFUEvlQqSZLFW9LuNtQoVbfbVafTMW/1+PFj8+4eOjLxeARpu8SXdeS1Ws0WTWEs+v2+vQeeB7aaUmG8LMYxnd50s12v11YE44UWkstDYN4Xb80zS1uERBzt42ufysI4suwZHoNCHRRDkiEB9igErQTBZi18pVKxjEY2m9X5+bmCYLNNOGlJlCObzVrHnmq1qmfPnpmn9SXjvKsvKPJ1D4wxhpI0IvOJgeC9QAP7fI+vGSDskmSkJ4bUHyj8er22tC9zts9j+WIgslYgDpDGgwcP9Prrr5ujoKwaAv2244UwAtJ23TYThcV89uyZrf1HadmQgd8ZOM9qk2bEisIiQ2xB9Hnmnp1kgKt41jt37ljjSOIxGkOcnp6ad8IDEL97VEFIwD3r9bqSyc0usiwsYaUcz0c248mTJ0ZeHhwc2Lt7hAEBmEwmNRgMjDlHUalY83Xp1FzQ5hoPi6dDoWaz2c4iIoybX4Tk+yd4ck/aNvAky4IxYezr9bqtfWDX4sFgYPeUtt2aicUJoer1uvXxZ/wx1FwTRIWhRU5wBCAfH1Z5wg4jyueSjCTl+XzfRghNSVZ8hdGkPJq+jn5xD8bV30/atrX3BUugNpwK13rw4IG++MUvWh8LuCwM8U3HC2EEeAmsGgwu+dD5fG6sLXlWjASTKW0HixbOJycnBj0ZWI54PG5bSbNj7n499uuvv27wHu/P/QeDgZ4+fbqjCKyUQyn4G3lj73H5js9BA/+8IaTklEUjbLXGtfDshC4YHO8hgIl4LF8PwXsDe0EQeFvOxTOR+uP5iE/JosRiu92OMKiksSjWgWOBUEOY6btHDE4qk8wFHMxisbB+ixjr2WxmZbhUN2LMQD2MCx51tVoZ0QlE93LijYa0XU7u43ycly9M2x8HzscZeF4Eroc6BWm3+hS55NqMKwbE8y+r1Urvvvuu7t27p/V6bcj341qMvTBGAK+FQPj0FnCJlXQ+xqUEtFQq7ZAqTBpWmLXshBR4J5QaLxKLxSzO5jg9Pd1p4dRqtTQYDKwoiclirzyvbAg3eWgMBTF7vV7X2dmZxeRAZowLwsU7DwYD5XI59ft9QyMIO4aGnXni8biFBtJ2ExIO77F8kRXVadQuYGwpTgJqQ7ghqGQ0WEC0ryRRFBlbjndGoVhyTG0G/yeTSVNqVkxWKhULT3w1KQU0+XxeFxcXlsrF0EGUgXS8gmNgkSkMNAYIRYOA9hkjjKgPp6Rt6IfhIHsBamA+CPe81/YkI8/AeXBdjAFzxdgvFgs9evRIb731lqGnjyMGXwgjIG1hE5OTTqd3WmGz8CSd3uziUyqVLA5l0vwkMCB4LeJNv9acf6vVZuvs4+NjNRoNq/S6vLy0RR3AvHa7bZwBcSU5X2IwH7vhKUmdwdyS7qFHIbUNwPLz83PFYjHVajV7DzwjuWsECTjJcyM0eGbiWg6MLudJ2x2BfRbDp/c8HCW/zZizRsLXx3sijm7IPizDAHA9evV5I9jv9w05YAAxemw1R88J0oSSzEAiA2xb7pEkyIFUJmSk53Eg6HxFKOd4EhSyFyNIuMZ5yC7pOlKKjKs34h6loRceDRCiYZC5HqlNnxL+4IMPdPfuXTWbTdup+KbjhTACCDJe2wuQJFv5hnLBZvuOOigl18NCArXwqIQXsNx4tVqtZp51uVyaAVqtNr39JenDDz80NMJ98ZIe4kuyHC9EE8gDb4LytVotXV9fm+chLUT6a7FYWIccFB8IjPempp9rAut5RyAonoQx8lkKv4wVRIUBwaDgGaWt9ycM4G9RFJmSMX6ENel02jIGXN8fbO+NIPt26+Px2EIiL+hBEOj8/NxkhCIgjAncDMbMF2th2DHykqzcmnEkS8O4IY8cjC3Gj2tioH16j5oPxt7/zWcpQFkUKzGH8XjcFllhdDyJCHqEzF0sNtv0FQoFy8LcdLwQRgDIhMUnxworDQTK5XK2gOP6+nonVvWVYT7OJP1C6osiIu5DOzIGGEYaaHhwcKB4PK4/+7M/M6QC9PJVikw+4QbvgxHCgHkmWJLVOnQ6HVUqFT179kxHR0em7LDfeEtiRZ6XdelAROmj/fS8dwFlSTLBxTh5OE8Nwj468ClODowd3hbPiQL4tFyxWLQNMVarlRXtQOLCFezzGxCq3KtarZoxGAwGqtfrO+lZDHUURVY+TBjAeIHa1uv1zv59nuj18+m5BGmbFfBhg7TdtgzPzuENEaQpSswYw8NIu+3MmC+MPTpDCA3S8BmQMAzVbreVSqX09ttv36p/n0Wj0UeShpLWklZRFP1cEAQ1Sb8r6b423YX+dvQJHYfhAfDUKBaWECtXr9fVarV2KrakbRcZrDgeDEaeSqpyuSxp46mPj49NAM/Pz01o4Qco4Xz8+LHVBnjF4Jk5MFpMwHQ6NTaZph++YUUQBFY1R5xII4h8Pq9arWZsu/dUMPYoTTqd3slKYOB8FZs3EAggnph4dDqdWoksHsu/J8YHIcZIcS6Gi/DF17LzjzUM3JOqt2QyuZPf51kg9nz3IN6F+B1vn0qljLz1aczBYGD7NCIjpDZZmuvhPEYApfahAaEgDgYE5YuhMKieMEQeee59783hHRuyhVNk/qRtqTWrN+lJAFeGTuRyObVaLTUajVt177NCAv9WFEVt9/tvSPq/oyj6J0EQ/Mbz3//j276MBWeS2VkHK0vtM00mvAL4sk+Ej7jTDyaK0Gq19MYbb5hV97wBXhqPlcvlrAoR5fYxI0YI0tDXHiD0/lkk2eeeOQbS8w4Uk2SzWes0vF905Bfz+MUvPu4GebDOHEGVtptuoNSEN8TAwFQMmkcbMN8eXvtqQl/sw3JpEJFPv2JMvJBDlnEdjB7elSpQipcwFJJsGbUvGebZIR+Zq2QyqU6nY+9ICbL30BgG+AoP4Rk7kAdyzHj5BTu+bJh5gUvCmTH2PI8Pb72cIR8Qv3Qf4r6+3gSZCcNQDx48uFV5P69w4G9K+qXnP/+3kr6hTzAC0nbNM1YWgfDFFwgzwsc5WHOgJwpBlZYkI4sg5WiRRSGNh/RkFICgng1GaLkH3oH4V5IZGe6H0nuYCmTmfRFW4jm/n12327UMCIKDUiNAeHCv7D6W5bmIP1ESxtR7d+Jp3o80LalZjIuv7OP9MOogERSMsfUbmMIRwKFA9FI4BqKRZCsDvQHgeVBAFIrUH+3fIS5BGmxU6qv4PIGKouKZgdzwSrwrc+H5DcaA8wkVPMHnjS8ZJWnby8LXC/iaCuQFNALnhKxhCDwpzN9vOz4LIxBJ+r+CIIgk/VfRppX4YbTtOHyhzX6FO0fg9h1gHTiKhDIxcKAB4mCEj8GWtqvdPD/AoEkbA8Dqt2fPnpkhYUUWNdf0EUAQ8ZDEsBgW4jPgJoYIofFIAc8G5PdVbFEU6eDgwPgQvAEkmk+XMg7ShhTDsPCuPm7lnp4UY5x8mhXo7QWScYP5x7N5YQblUHNP2pOai+dzvFNey7X9KkIEmjHzhKon2Fhcw7mz2cyW1WJEGQfCDYwRmYx6vW4OwKdtpe1iHop5CH1ADpIsO+OVi3DVr7MgreoRlXcaVBF6ctaTsL4uYh89gJhxjJ1OZ4evASV5ZwL6uu34LIzAvxlF0VkQBAeS/lUQBO/6D6Moip4bCO393fYdqFQqEcLFRDO5TCyIgFQT8JGB5lzv+ZLJpMVC4/FY3W7XhJQJo7KKYg0GHmKKRhcYA4SFxUL8jWeXNsroK/qw+j7dRoiz3ybq+dgolUpZOtKf6xUGZceLg2B8PAts5zzew8eNjJlXJmAk44yA4bX8mCPoKC4ejHvREMTPEULP/eg14EMZdtpF+CkNp5oQmE68L8nIUlaiUmvguR6MKcVHOArmiJ8lWQrRZxWYLzy9z0L5VDXj6McbYtmnCX1Kl3nwnJfPzkgyfsI3z+E+yAIoj3fZR4T++NRGIIqis+f/XwVB8HuS/pqky+D5/gNBENyRdPVx12AwE4mELcThwfEGXqgQtOf33YGgKBAscyy2WVDS7XZ34CCejMlm4Dy7TP6Z5/ENKJgglrd6Zlra9qYnzkT48S5AZWAzCsZ3IEcRGJqu+tCE7/l43hNSIAi8EOMFBPZQE2QAt0Hs6VNcPtMAr+G9jScRi8XijgHFg6JIvhjKGwbeD8WJxWLWl4BK0tlspnw+byswQY4Qet1u1xzFYrGwtSZRtN22m3dibpgPxt/zSMiiDwm9YiNreGAvk4yrV3SITF+v4slMD+s9UuH/INgsCoI7k7SjFxgi5NSTkjcdn3YHorykWLTZkDQv6d+W9J9L+t8l/fuS/snz//+3T7iOgmCzgKfb7e6k3BAKbwTIuUvbrACCE4/HVa1WlcvlNBgMdHV1ZYUpHiJyLWm3gzBeKZVKWcsr1u0j0Agf0BIhRugwHqlUyhqoMol4CyAy1X0IMl7g6OjI0oL8nRJoyEWMhkcCPpb1K818jp9qPpTOL6vdryhEQCFpOYhHad3tmWzOJ2b348OzYfj5nVWQnh+CFKWen3mmozL1CMvl0tqlo1hhGFqrM1J17OvoS6fhfoDoXtm8MpGe9QaduaSGwzsDaZue9YbF8wTIHEYBGfecDc+DEoPoeFf/rN6o+LDYX+em49MigUNJv/f8AROS/ocoiv6PIAi+Lel/DoLgP5T0WNLf/kEuxqD4XLh/SUmmENJ2QCCRWLYZBIG1I/MVZ9Iu+8y1PYz2qR9pa2TwCNPp1IQJI8A1fCzqIRif4+3xbvQJQHjw4MSwoBpJBmGJWT068ZkBuAdfZ8F7eIgLU+/Zf++teGYvRB7mw8/4d/RzBhsOqYin8p+R3vSKA89BLL9f9egVzaMb5gvEwApUSZZ25f1YqUhV32q1su7IPJ8nUMMw3Ok9SHGWTyXyvKRIPafgjTtzxPMynvzP3wiBpK0BwJDwHjgPb/y9wcHYMOe3HZ/KCERR9FDSV274+7Wkv/HDXIt95/BwKB2WjZfwlhLvxTprIFmv19tp3CHt5malbSaCAcRCS9ulxLHYJhUHk+yLPeAL6HeHoPhtuShX9SlDFAzDRMUiuevxeKyjo6OdsAHoTPoP4gzD4mG9h/c8Kwq/j6Jg/zmPwxOu0nafP0+8Mj6kV/mdMAFD60lflIwGKdwLAwJqwMDGYjFbCQh6w6hhDFmyzByxcQweHg4DxfEhJXPKzz50wjh4Qg3eh3lDUZlTalw8qerHzDsI/+7+Z97bjwv35n8fnvjQT9puzgJH4I3/bccL0VQE5tg3WMTKSVuW2XsyIHCj0VCj0VAURer3+ztFE/vklSekfMyJIUDAKOChkIS4l4UjEHWkZ/B28/ncCpMQWLwKP/NcVMUhMPupTSAm3/FWHyHnfkB64KqPB7meL8WmU5NHR4wz//v0Ks+FsGEEfOk2/zzrjbf35Jsk4y+CILDwiXDOE568L6EhYwlahCzGIBMq7Tf4iMfjlnpkTH3xjecFeD4MoSf8CNlms5m63a69n0d6vjBov6QXpfZk9j5ng7L7efCGiWfmn3de+zyZn4fPOzvwqY9YLGY1+wgRCuvJLWnbybZarRrkBfrzPQQEJd/PIPA9BhchQNC9p/OpNs5LJpPqdrumFFhjnxuHicZoAB95F1AM12N1XbVaVaPRULlctpQUXZZ5Nog4v6DIE0I+fIF/QEi9h/NMsvdEpEe5FukoBG/fo3HfMNz2GURxS6WSlVf7cACl8DAdwyZtOw0hF2RuwjBUuVxWv9+3zAPvKm35HQqDSCd65tyvDfAKgrEi5vZj5wk2CNvpdKp6vb6zZgHDyNzjsDBMvAMhBDLO+Hji14cJGBOQh0cgnku4yaBLPwYtx8mdekH1kMenZGifVKvVrLEnA+TTWT7eRaC9QnM9PBRxJN4A+AohBKS/vr7eUTxJOxaaQec+CHsymdxpITaZTGxhkIeNkqzlF89JCozUqU/VFYvFHeOFQFCL7z0GY+BJKmlLYHlY7okvID7IAePnszYe2nJd7ufbkPk59hCWcUT59/eBAN5TWISnj8Vi9nc4E9KHpVLJNoAhZOA9vIJReuxrUTy5J21rGzBgdEhKp9M6Pj7W5eXlDqHN+Hs59qlV76k59kMPb0Q8emVefAqQ7zBvnhMAbd12vBBGQNolqyTtEH7SRijK5bJ5RHbx9Us5fcWftI33EXAGH2vPpKFg4/HYPBmTSLbAx5D9ft8sMOQTlYJU1h0fH1tP+FRq06aannqQggg9XoRJ9W3FYb4nk4nVJeAZQAlscc64cT/CBo9ovGJ5gSLc8saCvzOmfm44h3Ek3QfDzzjueyfu6TMmGEAWQ7FTL+k/ajuy2axOT09tQw8WItFZCOiMvLD4imImwiqP7LzccV4YhtbCDcNLuCbJuCvQHhu/YoiRUU82Mu4+Zvd8jbRbLegRCmjEV2b673Agn/w9DMOd8vTbjhfCCOyTIZ69D4JNtxm/AywVeggSE4qF9zERhw8HiPEZ7P2KL7yhJCPrqAfw8S9KXCgU9Morr9h1SqWSGo2G1bz71tvkdoMgsO2yqK2XpEqlYmsFmMjhcKjJZGJhAUpItRjXQlgzmYzdH6OAgkDKeUHDg8At8K6esaY6j7FcLpcfaUiKEuAp+bv3vrwTPApjTAi2XC7tff14cw5dgOLxTcfl8XhsRPBoNNLBwYGRrNfX1zo4ODCFYQ0+MgBCpLWUFwAAIABJREFUpL0b7+0NHiGBpB3l9aFZp9PRq6++aoVuNxVucU1+9mQrxsB/hkPwaM2PIaGAr43x6cN9UvDjyMEXwgj41BKwS9q8+J07d2y7pna7beczKJ7Y814PYfNhgq/4Q5Bo8AFxRM0+cfFwODSSECjabDatDiGKop321DDjtB6jKzBVXoVCQcvlcif2JVZFSBEAv0KNc2jSgZHw26URi/p25zQq8b0Ofd4fYUR48I4YMO/1CbsIj0gr+hQf191HHhBl+01dCBVotgL5xtJe8u8gnkRi06YMUpY9C/Gwg8FAxWLRvLW0zSQQmvhMCc+P7MG3eNnyaVjvAKgn8f0k6WDlG68iy4w7yAc59nUDzL8nWBlz1lbwXHzmOQ1vnOB3pB8DI+CZVWlL7hwfHxu0YnsnFNnHbL422w+q93j+HlzfKyIlv6T9/Eq0k5MT22GW7cXj8U3rrvF4rIuLi50VjkEQWGdjGmvS6ZfCI54nlUoZaciCITwysB80wTm0YIdd97XhoBnvifDEniegnsCjJT731XDSNmaHvER594lDDAjGC2jqU1g8I/X1CCy1F2EYWiPZVqtlwlutVm2dQiKRsApQX64M8qGpBqSdr6ak/Tae2ZcVM44opyfeMEYYZD8/zCX3ByUhH76wzZcDS7uZmX3CFcPA+GMgPSfAuRhpH8KCKpjr244XwghI2z35EomEXn31VStDbbVakrZNFhFunwL0sNZbSf5hIIhvpS0ZiTGhTXYymdTx8bG1q6pWqzulm51OR+122z73oQkKjifBOAFVgfpcC4hbLpd1dXVlFXbSRij2O9giwJBmfOZhIWXRkqwjjYe4KFwsFrMWbRhFX1tArQRz43PmnrTD8DKeHq76WgZvpIIgMNRCbM0qP7oNI7hkfzB6cAbZbFZHR0f64IMPrNKU4qhsNquDg4OPNOCAE/B8BP0LUWCgNNAd44sD8itavQxSROZlEZ6Bd/fEoLTNgPAdb7w9se1JXM6B48IoEFIzX8iED7VuO14YIwB0rdfrVg56cXFh3tIrlbTLiPLyPk5igBFKICBpOphioGA+n1ej0TCDwVLaq6srPXv2bKeXG5Db5759Ggf2Gm/nKwzxoCgOS4h9i2hpu7AHQaBGAeEEcfA5PAfjARzk2VBQ4Cpp2fl8bm27GF+eyVdLenLV16T78Gp/VR73xxDy+37GgtCM8IZ9HBlzCr7wwNfX12a8MJyER6wvINRjD8coilStVnfmggIjDBjhkicqkQ8fh9OOjLH0NQrIgYfhPsWKUfT5f49qvTPzGZT98JY5IazxaAHjgDH3hutG3ftRFPazPvDWtPI6Ozszj+nhv/c8+4yzT9N5MgarDnueTCZt8ugchLUknn/w4IF5FrwQQumbQrIVON1uEQiIQH5nYv2GncSPiURC5+fnO5wD0A5PzzZYvAtx/ng83qkCZMEUBtIvlqKOwXsZvP96vbZno5IS0g9lQMAhFxlnDCGeGsFEGLkvPAXf8+Xh0jYeD4LA9tXjnUqlknUPJoSLxzct4999911Np1MVi0XVajW7F6lCvxMT7Dnjxv0Jg1guXqvV7J1BZn7M8M7IgU+dMl4+pQoCAN0yTtJudyfPlXgEi9Hy0N9ndnzGBWPlQzqPJG46XggjkExuevzP53PbjAPFRUiIBX3KBTjm01C8MALrCS7iRpqLAI9ns5nOz891dXWlWCxmtQcItqQdRfHkYxAEVjaMQWKDEt8uHU/HhhnUHHB/Jp9Go/yOMUIxpC2HcnBwoG63a/ltX2GJ0HrjiGdAcFDSQqGw0yBkvV7bfYGevgALbkCSoSkPo8MwtPQYqAR0heHwMaqPw6NoU2BEc0w8/dHRkYVU1WpVV1dX1jSE9z48PLS9A+v1ujH1fl8DFCQIAuN9MAJUgsJTMGYQtpJ2vCyKy7uydoBxo0UaCO2mkNWnDuF4QJDcx1dX8gx+HYUnxpFXjAEl3S88EkDoHz16ZNCQCfED58s5mQBp2/11P52C9Ze2a7C9Je90Ojo/P7flyz49iSf3MVo6nbY00H4VIevTW62WQWq/7x9sfRRFury8NBTCZ5A4/nmXy6UqlYoSiYTa7fbO70w8zTZ8TC9t6xZ8PQRrGRhDT2Z5YeK9MXQ0q2QumCNQA23Vut3ujrejM5MPWbyQMlcYbEIar0j+XUj3obCEEoSPnU5HZ2dntvMvSK7ZbJpBh/tByZh31nBIskYzk8nEDDgywwKlbrdrm+HuQ3yMj98pCwPoqySRXU/ygVb86lj+BxXuh4p8D5LYE4cQhJ6Y3T9eCCOwXq/18OFDmwSWhfpqvv2BBnb61AcDzKR6YobKujAM9ejRI3W7XfO8WMp4PG7pN/q2IexkEFBswgKIJZSYjVAxHLlczs4FikJgcV1KjWGv2Xa8Uqno+vpa/X5fh4eHO2sX+v2+rW3AA/NshBKEEwgJCo7n8Wv7qYz0G4t4gSa3H4/HjaDM5/Mm6JJsjwS8oE9Z8X2Unp8JxYJg2/STsl6arC6XS2v7LknNZlPtdtveazgc2vbc8/lcjUbDEFY+n7cS72QyqV6vZwgIchdUwnsT2kkbY9VqtWyvRGo9Dg8PDbGiyD6bgxLurx/gmr4KVtrdrNb/3YcCXA+FRg9wMD5UImQDtb7w4QDKT1UdgiRph0gDFiHMPh1IjIa1J6WDt0UYr66ujBC6vLy0AVqv12b5gYMoK3AP6OnRSaPR0PX1tRkI4DzLmjEilJnyzHTW7Xa71v/g8vLSEAfjAqm5WCxUKBSMkPK9+SD1pI/ufcfPPs70OW48G97dtxrzpKuPVf3+B0Gw2Vw1kUhYO3hvjFqtlnlyz5JjwNnHgcxKu91WrVazDUR6vZ4VSb366qtmoBuNhobDoUqlktV1vP/++2o0GprPN/s5gqZms5ltYAs3A3qiNoHwkPMxghhHsj+sV8CY+muGYWiGkPfF4/uiKf+Zd2rAd8/2e6jvswfcGyPBO4AsfSiIzN12vDBGgBZUPpePYqOQDKgki+f8ZIRhqGazabv44FUYLFJQUbRpDOK9HDEv9QgoAOz5fD63uJlnZpEJxgkOAsXBs/nMBvsk+jTf2dmZstnsTqflRqNhi1TCcNMgA9YbRfeKz3bhpMX8c3hj4nkN3hvj6Zf7+oIVXxTjDSTzhZFqt9tar9c2/uwczPJu4lS8p0caxMGEFdfX16YU9Xpd0mZzkjfffFPHx8fK5XI6Pz+3BqzIAAVbUbSp76AFGWEN5/kGrvQHKJVKFuohW+wR0W63zdjeu3fPWpzBERAWEjp4xIUT8MYVeUHOUFrIZByTtM08+efe74UIWvDXYw2In9ebjh/ZCARB8JY2ewtwvC7pP5VUkfQPJbWe//03oyj6g0+6HkIqydIavBx/3ycGnz+H0um0pYJ8aTEWHaHDe/pNPSD1sKhAcYQH6woByGRLsqwBi3n8KjKfEkPRECgyAfV63ZbCInigEWLI0Whk3n44HO6gC96fscJb+9WK/t3gAHzlH+PljSww0pfOYpQJMTBECKMP19gYhkU88A6++hFF8fwOKUhJ1m49ldrsJZBOp61tGNWAqVRKd+7cUT6f19nZmc7Pz3X//n0b12QyqWq1qmKxqNFoZBkZwioyOYeHh1qtVhoOh6pUKjvGGAfCeEqbXaOOj4931gqQScBwEhLgpX2aEHlnLL1cM96cu58WRLExLP47cADIAWMAIXzb8SMbgSiK3pP01eeCGJd0Jun3JP0Hkv55FEX/9Ie5HrENuVqEl4H3uU4GJZ3e7EtI+or8PVYRj5vNZq3Ax69RX61WlkaDRPLf9blZb60RikajYev58UaJxG7HXMIZ1hXgIfEgUbTZtisINrsq8RkrFTudjkajkWKxmO1biMVn7z2f+WB8PC+B9yPE8u/HtVBQaZvC9MSUH3eEnJiYqjzid7wPvAXhTr1e12AwsAIrrgmikjY8AOEa7+G3Avvggw/MML/++usaj8dmiNhJCqPFbkenp6fGC/F8zBvl3oxju9227BEIbDab2a7GxN6PHz/WG2+8YUqO9yZjAWIDEVDkBBJgjHl/DAHGkXDXG3feCxlFNzAW1KGEYWhhHrL+l1Es9DckfRhF0WNP1P2gB4pF/Ip19ddCGBkM4DPC6PPc3pvE45uNLDudjg0EXIHvCAxcw/NCIuFNuR7fRzGobiR2pXAG7gAPB9LxMRybpGLhSR9Kshp5SoqDILD1CrPZTI1GQ8vl0jw7MSexPcrKu/rKM7wQwkk603sjBI3vcp5PM3p4G4ahPR/vzFr/TqdjOftKpWJknkd1hIHValWxWEzX19eG5PwGn4PBQL/0S7+kd955x0KPKIrUbDZtjYnfvj0MQ9vHMBaL6eTkZGcVaBRF1pvAr/nPZrOG6Ai3yIxg+B49eqTXXnvNxkKSGQzf/m0fQfmUoaSPrNPwBDdzwbkgE88t+M8zmYyR23z34/gA6bMzAn9H0v/ofv/1IAj+vqTvSPrH0SdsQUY8ClEEFPIr6XxuF8g2Ho81GAwM1vnNKsh/L5dLS6tASBEb++o8lBRvBdnExIMiUOh6va75fK5ut6v5fG5bp+P9stmswXLv0TwxhfITH8fjcY1GIzWbTRsLkA1LkIGPfjt20BPwEPLKv5dPRXq4KW1JP4yIh6C+voGwwe+J6FNZ3qMxdr5RarfbVaVSsVWh0+nUxg6jxQ66x8fHOj8/1+XlpTVYWa/Xun//vr73ve8Z8YvnRQlI25FqC8NNWfZgMNCdO3fsmZEXlJ8dj30xVCKRsFSkLxbC0KdSKZ2fn+utt96ysQPVoMi+1p+QAsSIU9pHAhCpeHd0xJOE9DLE0BGOIN98D4Tm+bT947PYizAl6d+T9J88/9O/kPRb2mxK8luS/pmkf3DD92zzEWJgiBnYZzwmE0FF4Xq9Kfn025IzGD7/ul5vSnZ9M08GCFiHMpIVYBLYugvYDdLAEFDTgBX3nhWlIh5DoAgJmMRut2uGarXaNgjB61Blh7E6Pj62hUpeSSEcKVf2m1ryDNK2jsILnOdhfLEPz4jnkbYt3zzKIJ3Lastms2ljGIvFjP3Hu/K8hUJhhwz2oZcvnoFT4P7z+VzHx8f6yle+oj/90z+1bAw9FEhZXl1dqd1uG0qRtoVkPtQAzbA4LJ/P72QP+A4yA4FJzYUkPXnyRHfv3rU9DyANMRQYMMIlxgNZ8usLmB+fDcBwUcjlZUuSyaq/LoQ6SOHzzg78u5K+G0XR5fOBvuSDIAj+paTfv+lLkdt8pFarRRRilEoli3cQ/ng8bl14YPg9OUahDNY1DEMjg3wNN54J7wYzzbJPznv+TJJk9QqQLCghdekodafTMXhaLBbNE0kyQ8bP5XLZmlLCtPt8si+bpZMSuXf4AghOaeP9yGGj6N4gAjE9ew1qgqfAuHpii5LqXq9nAo1HYwzX67UODg4kbVb6+ZgX4+frFVhkQzYDzzoajdTr9WwhDtu1j0YjK6y6e/euZSkeP35sa01Wq22jl0qlYt/3cTJjyHmkkcfjsXK5nHUjwptGUWQ1BZVKRdlsVtVq1XgQDEQYhraACSNNyMB8cz5KyVyDZriOzxjgSAitMGYYAD7z3BmGgtAJLoB5uu34LIzAr8qFAsHzTUee//q3JL3zSRfwRBUvhDAfHx9L2lhxFpYQy6OgLCbBGuKd8Rx+sHx5JhYYy+u3dMJrM6DE75JMIHwVHfcizehjTE/OkeuH6KQICeKJJageDS0WC0NDeAzGiZjXL4MllOIgrsdg8TsCxvjzMx6c7AEFUNJ2MRBeB+TAHAE9WVZNKICQ8vysxSgUCqrX6zuZoFgspq9+9av61re+ZQiD8aXu/9VXX9Ubb7yh9957T7FYzGoIlsulDg4OLO0cBIGePn2qarVqBVU0bfHFaTwjRU6+XgUPTk9H6iDoKC1tHNHFxYXu3Llj8+Q9uS84Y/58Ph+jsd+GzdcCYAR8ZSljieEC6fqaBM9Z3HR8FpuP/LKkf+T+/F8EQfBVbcKBR3uf3Xh4eIMgk7eVZHlmct8e3qzX222rEUissSdZGDhfOw+hmMlkbHcf2GP+LsliV79KD+bYp88wCoQhxOCNRsM8r0cANM4Iw9CEKxbbbsUG4YhC+ko3ipgQQrIDfrmqTxnuCwHKDkHqF7kgNH4RE97bl1L7unsYf8IUvyU7QonSgT5AcsTiePXZbKbvfOc7evPNN/Xs2TM9efJEiURCT58+VTKZ1Ne//nWb9263q/PzcxN6UFs+n9edO3esgpDr0gSW9wLhMH/wQTgXaaPsLBLjHsgITofx7HQ6tpAJ+USRfYYJGdpHaBgDnJbP6PgCIp7LE7o4LK7jF3p9LnUCzwVvLKm+97e/98NeBw+E5WORTxAE1kSDdQA+hvLxnreMKDtEDd14/DmSLHQAEl5eXlpVIF6F2J8iH0kG96Nosw6AuD6dTqtYLEradq5F4fHmlOZKsq20gMjlclnz+dzIMRYO4b1gvCED8RrUPQA9QQ8INu8NGYYQxeNx40JgrUFGdFPCm2J8b+pZh2Iwh75cGy+FN/OEFc/htxnjWCwWevDgge7evatGo6GHDx9aSvS73/2u0um0vv/97+vXfu3X9Nu//dsGfeFuer2eZXtOTk5MATHI3W7XDILnIvYVMhaLqdlsGv8DlwTaItwZjUaqVqtmOHEgfmEWjgDFRf6QY8bpJjTgCUNQFfPix94vLGK8+ey244WoGESg4vG4XnnlFfs7a/6Bz8AhXl7aFrb43ClWk3wplV9YeCwmnp80UK1WM0ubSqX07rubvVUh6BB2vBrFPUBkQhXPjksyuI4ApVIpa4nFMzPBCAQCT8qTTrp4HZbKklqlGxJMMQaAzIFPP/kUJ41VKANGgPdjf+9RUDSem7HE+zMOVN35+gNQBOMHLA7D0DIBsVhMr7/+uvUQfPjwoX7hF35B19fXunv3rt555x2bdwhlsgQobrVaVRiGevLkiW1QUiwWrRXc3bt3rZQbVMG8Y8AWi4VqtZqhSlh8j4JACDSASSaTVvrsVyXCCXiF9iEcyBR98PDfczQ8pyfCkVlkJ51OW3EUxOlfRp3Apzrw/sfHx5afpUrMV6ohxHg84liEyxMgHikwSLT4okrQQy8gZBAEOj8/13g8Vr1eV6vVsrgYRRiNRsb+Qg4i0H7xCykvv1tNGIY7pczFYlGZTEatVsuQCGk6lIO0E14FY8J7+7JQ8uxATUnGKkMS+SwByl4ul21JM14JT+cLWgin8IZwJft1CIQWEL7Mj29QQo5e2oZ1pGbpDhSGod566y1961vfUrlc1htvvGHlwKz0Yx6ZI5+GZTOS+XyuZ8+eablc6ujoSCcnJ4qi7fJhajt8tgBk4hu4gIyoFZC2IQ+cQRRtloCzZJz39Z4ctAqS8FDfZ0yQb5/mw+ntF74xNzgg6hTQsduOF8YIQKhcXV2ZAQDC+Th1H/74mEnaKj8FMBgFYlcGi5jdLwuGHKSElErCxWJhqadcLqdOp2MNLYHRcBTtdluFQsFgvKSdohZKjTl4Tiw19wa2o+zwBvAl7PpLtSLjAi+CwHBdkIqvcYcMQ2nwZAgT5yP0eHGfWuMz2m4jeH71IJ7TCye8ie9LAHG4Xm/6GfjQChLuG9/4hn75l39Z8XhcxWJRf/iHf2hViRC5k8nEsjXMLyllOJk///M/t8zLwcGBVfklk8mdgifGxmcyeCeyBSgscB7k1e/3bW7gsUhH+zUNPiOAwoMAPCqQtGMQMADMB3PGNXEeoMLbjhfCCABtOp3ODtnlDYAnRCB1EGYUkMHAyzCJ/nvk4IF9TEC/31e73TbDEYahTX6xWNR8PrfONp6LgLXH6+B5SCmiVCgskN2TkngRhB5B7na7mkwm1i8viiJrtkElIek+SUY2StrxVIwBUNYTgDd5GAQPIwQiYxHVPtcgyWopWDyE0fXcBFkXFIfn5DwMOyW1tBcfDoc6OTlRrVZTtVrVn/zJn1ijkfPzc1Ny0Atjul5v1pAQUvlwAeW+vLzUZDJRvV5XJpMx+E+cz3yDzpbLpe2AhEHAi2M8CZsmk4lqtZoZQzw/yo8zwHggV16pvQf3qHjfIEm724/5mpj9ufqI/v1oavvZH8+ePZMkGyQEgoHgpXwMRWggbTviegJkH0HwOR4PiH1xcaHr62sbtGw2a8Ljl9rO53N1Oh2Dip5AohqQ9BMxIJZ5Pp/r5OTE1iiwChKuIBaL2aYjKATKWiwWdXh4aOkxClCoeYDXQPgwIj7njxFh7DxvwlgBWb3X9wLO+PlUHkKHIcLAQbYyd8BXn8UgTMFo+AUyPA/IAiP01ltv6eTkRO+9956VVg8GA+MGTk5ObOXl5eWmZAXYT+YCp0E5MWtL8O5+QRBkLsq9XC53UCByACdCwRVIazKZqNlsWsNcz9L79KqvDOS9PY/DOV7RcXYcyCNGAfkGDd12vBBGAAFDIRhAaZti8WWWxE3eu0kyGASJSLkwRAyKgqDScCKTyahSqezsZ4+ir9drnZ+f6+DgwGJbWoN7gfJltEA+7s01eXbitGQyacUw9XpdicR2xyU8PmvpKZEmXkYoWcmHB/V1BRBN3otJ2kEPvobBE5QUs1CC7YkocvYQqzwDnAL3weCh8D4vDrHmjYFfFcdYSZtYH8PHZi1vv/22qtWqSqWSrq6udHp6aqXGw+FQX/jCF3R6emoxfKlUslSkj/ExdtImJDk/P1epVFI2m7UNb3x9Aw4BdIDcgdLgJzB+fpUlc4aR8Mw/DggvzjzthwMYcb+EnsMTmn5tgbTtznTT8UIYAWm3eQiKI23TTwgfg+RJLw/7fToEoaVyEFhKWMBmJigWBgcloo7ft5Mmny1tLC2eot1u7ywOQXk9MUMhCptn0NCS60AiwjhfXFwY6UYJL16fGoqb8sc+24KwMq78Y4WjX4kmbdKWjKknWSH0PPkHX7FYLDQYDCRtOwJTyowQA4EhFIH8HB5d+RLbdDptKz2Xy6Wl/uBXfvqnf1pf+9rX9Lu/+7s6PDxUq9VSGIbWNs7vZ4CMEUcXCoWdBT4oT6/X02AwUC6XU61Wsw7IQHtKhpEpQkyPkjz6BB0wL8iZV2Bfxcr7Mb/IvDcInO9DU+aIsWbeCDduO14YI+DjVs/aI+S8GJPoz/doYL/s0sMmrO1oNDLDQMEFBoK4klQeXWn8xqQgEV+SCW/AhC0WCysLpZiJXgOED6lUyoSMKkg2K1ksFrq6+v+oe/cQS9fsvO959626bvtW1dXVfS59jg8a60aYiUxsCHFMnD8cE5g4BEUmxJIjQgQyScCQSI4hIWAQJE5IMDgoWEgCY0dEji2CIJFFQhzkSYg9Ropn5sg67nNOn+6u6967alfVrsuu/eWPXb/1Pd/bu7p75pwZVT5oumrXt7/L+67Ls5613vXuaTKZxIIkFB9mnHCF8crJPtCIFwshTL6TjaQYGycCEW6E1tOAninweN7DLrIgnMs8euUmz+0cBEaAbdV89SJz8Lu/+7s6Pz/XV77yFT1+/FgPHz6UNK8wffLkib75zW9GmzHQEdfo9/sRSnonHt4deWSNhjQvIweZ+Xh5IQ5Gl/UveHbkSyp5K7JgHg6gwL4oy5WXNKHXcywaf77jCOPOcwIuiM6YSuWLOBxC2D3+dzTg9QQcHiJ4EQWTSEHSaDSqlAhDIlHKe3h4GM/KSkGMAQtjrq6uos01gtvv92P9A8VGeBaanVCtRuqr0WjEoiLPiNDUFIiOcnGeVO7B4PEsz8374yU5xxXVU4EosqTwyCgz5zH+7u24Ds/o8avHu6ToMGrOI0AOYpjr9XqUD5+enuq3fuu3dHh4WCmgQQklRcqVngJra2vBjzQa1RWDnmp2kg4U6K3X2ZEKwo/6DUJIwiVIX2frfQEX44Bcc0/kl5/diCxCy/zd5d5DkDufHZBKZfO0E0rug+ID4hPlZBeD4ylAJsPTcUyYCycpOgg2VpVhvX33HRTNYzBy9qurq+r1eqEMLELhfF8BeXFxUVmqjEIiyIQY5JOpRgNe+y657tUYU087IYSMJQyz167jJQkdyDQ4BAVmkqkBvjPePI/nud3L+tzhrWH1QSC+k7MXa7GIDNn4+te/Hs1Hd3Z2Yj09JC98BRkhPLxDdNKvvkKTe1MgRTdlQkfPUIHMQILIpHNbLuMeooCYeEdHScg053uth+sL1/dwGV2hjuC2484YAalUaiwink6qLjLiPAaAAZHKZbJ4fCeiyNG70jC4QHcKilA6jJFDaLySFw2hvBy9Xi9aUEnzgpPBYCBpvhaBlXndbrfC3KOMa2trkU701YH0Scir7UhRSmVY4IbTob8TTbyjG18XGJ7LoSrneM0BJCH38SpF3itHGxCReDUyLMwvtRYgRAwsxr1er0f+n0alQOx79+7prbfeCtSwubkZnYWHw2G0ekMRMRoeasIdAMPZPXplZUWPHj2qIDDG1TM4KaUo9XYj7OPqc+uIgOu5wvs5fMZcOULg+m4kHBXnx50wArnQSGU/ekkVy8bf8C4MhnsWBter3Yj9vP+Ap8wQVC90ubq6ioIPRwgILY1EyD9DiHkVG4w20JvKO54VD+XPUqvVojjJu924t5bKZc4uUFLZQ8DRCod7KalsqOIIAqOHB+QaKB4GwdNaxKpeMu1kJM/uc848cC7hE+GHr7Pn+iygoiy2VqtFm/ZmsxlK2O124zngRryepFarVfgNr+ZLqVyJ6mk5SOXLy8tY0MaycAySI9K8hsV5LVdOlNnH3pEA/zuH4oaBe3po5s7T53vRcSeMQB7veCzkOX33SnnJsFtUSZXPgVPOlKdUdm6t1WqhcExyrVZTp9Op7MQjlRttUJcNzOfvo9EoGm3SB5/UGe8IhAb+AfHW19fD63lV2NraWoQPGCeP9RAw3sULWBhfR0VO3iGATsiCKFBqFNmNJWOK0nhal/eRSoMNeeUGid+pGyC7QUf2AAAgAElEQVTEgB/g+qQkJUU9wdLSUgUN0dBjMploa2srDFur1aoU9/BMKBz8gz8zWRMUkvmCo5EUhUQ8P6XllD4TtmBc4CAwKs67OD/lkN4NBs/tRiXXGecYHDnnIW9+3AkjwAuQMspznG7FOfBGnkmQynjTmWyux8AwwY1GI+J24Cyhg6OTyWSi5eVlPX78WM1mM7zO3t6eJMWadxqdDofDiEclVWJmT+F4zzkgNVt8UWgE6cT3ST3ihREsDF5OAHnIw1hgiPJwwVf+uWdH4GClnWsgHAHeUzeA4DGHHqN6poK/8YzUL0jz1ZooIUjO112QbqQgBti/tbVVWQchKdZGwO3UarVYTOTZIRSIMAXld+RJGo/CITIY8AruqeFIIKYdsi8KDVwncpLbjQAGzB1GThhybc8OLTruhBGQqkuBGQyHlHhNZ3ERRBc2SVFdRm06nl0qBY94G08Hu8uAYvEfPHig9fV1tdttbWxsaDAYaDab6fnz57EMWCrhnjT3pBSVIEgQksSXQM7RaBR5f0qmeZerq6uoWMSQeVoIGCpV6ywwgu4JODwcwGDAiUgKg8iYYizy5yJ370pNKo70mM9RHq9yvj8TNRccVFWmlCrpxslkEpt8YKwg8CaTSfR/YKx8Y1ZQCUpPnQbemecBQbohwOMS2jUajWhPTqjY6/WiCQtoyccf9EYTVdAA78b3mGvCNVdw/911xmWE53+V8nO8kRFIKf2CpH9V0l5RFD9881lf830H3tO8eciPFkUxTPO7/jeS/qSkM0k/URTFP3zN9eN/Zz09BJDK0tbcKnqcSnxHLDudTiM1g2X1dBbe1uFat9vV9vZ25P+pi//ss8+iAAalJmXnqUwmmHc4Pz/X0dFRLHDxdeSksIjjR6OR2u121Ba4IJFqYqLd4xDDgnbcS3DgGT0WL4qypNYFBiGnuWkufPArrMJzAs0JTkIJrudr5Z1ExFgxh64YLFCiIo+ai+PjY/V6PR0dHWl9fV17e3v6+OOPY4v5RqOh7e1tra+vB4+AV/Q4msIlCFZCm9PTUx0cHGhjYyOekbGDqyGbIaliXDY2NiLLgXw5SiAMdGOJs/KwyhEVqICD350H8HDLyUOXg/x4UyTwi5L+iqRfts9+RtJvFkXxcymln7n5/T/WvOfg9938+8OaNx79w6+7gXt1j4NcMBkIr7dnIBxy0TyU7wP1ER4IHsgbhJY9DNjsoyjmS0I/+ugjHR8fV1hj8r5OZPIZz0I44YQWguKxJH0VeSaMGL3v8A7EwjwbHtAXLXnog2K6cHh+nt85EEaUUVKEaC5wRVFEPwOeCe7i5OQkzucARhPXe22Aez4nSHlu+A/6L9Av0IlKti4/PT2Ne4CSvJELaIDQk2vwjBgiHANz6S28Ly8vg2egGSnhAWhpb28v1iQ4wmVOnGzEMDL2HsJw8Nzcx//O9fOwAMMDyvrc4UBRFP9HSum97OOvSvpjNz//kqT/XXMj8FVJv1zMtfhrKaVuqvYdXHjkbDKfuWdzK+nw3qE468ghkFxpESxW9w0GA9Xrdb3//vuaTqdR6NNsztt67e/vVzamQEgRRshFBhiYDx+ARyHOpe59MplobW0tBP7evXva3NwMQ+CbqTab5dJWhIEmqhCmwHqH5h4meSaEscMYeAiDYWGc+Z43DsEYOFnoKIJCJhCQVBYYOXPOvXlOXzTlCtnpdGIthufhcQTj8ThChe3tbQ0GgzDCeHOQC56fmB8Edn19HbwOPMH19XUYWL5DLQPvw1hh0KQ5LGcF5GQy0TvvvBMI8OLiIhwHyM/RgpOwPnfMmfM/bijzlK5nYzjvu8UJPDDF3pH04ObntyQ9tfM+u/nslUbAoZlbMOJSoBewDYHBMmKRj4+Pg0hkMom3fRFLq9XSu+++q0ajoc3NTZ2cnGgymWhnZycE0llbPAY7AkPqsfEoO+GQR2YSgOq+CMW70fruOkBdqSR0rq+vo8NOv9+Pxhmk01AWh+w+Jnl5q3tzBNGZeGogEGyu7XUCy8vLsTCI83l2xh5CDIKNcSdDwxx7eOBhAvX5KAUFVp53Z9xZt3B0dKTt7e3KWFNDALoidHOHArkJ+ei7UmHw4R68aEpSXNPHnedqtVp69uyZOp1OpJlxXqAQN8Q+T64TGAPnCFBs0IovrmN+/R7f9RRhURRFSun2HMSCI9m+A9TE8+DOlnrloCMEhBurh+fxgYJZJ1Mwm81iww6KRSaTiT755JMgn2D5URDYel8I5NucSwoj4Tnver2uBw8eqFarVbrU0kffFTknJckMLC3N986jpJn1Bhg8jCKhgvMpKJ4LPM+L0oIsMEp4IJAM80GRlRtGkANkFjEzEBjlxWDDnHt6E9INwWfM3SuyrbgXCKHIZEpY54FS7O/vazgc6vnz53r8+HEU8XAPFJhGKLwnRF+OPAnzkCuKpbyvBdkGn8/pdBoFZqwRAYVAwDIfhKg+Z5Iq8+bcC3riIRRhlzs69MC5hPz4PEZgF5ifUnooae/m82eS3rHz3r75rHIUtu/Ao0ePCifqsI5YZ/dEeAyEFjKKWM7ZaDYzmc1m0RMeRFGv13VwcBDe48WLOVDJY8WTk5OX8ulO8GB88BoohK9UYzJQWG9nzbuQ53YFJV/9+PFjLS8vxz0oFXalQ/God3ABAgrilW/GP57FjRfGBwX2MArDxxwxTnjOZrOpo6OjUOZ6vR5ZFjgakIOXWoMg2A6MuSWsw4BfXFzEWn4UCPjOdfb396Ou4OnTp2q1WlFhWRRFrANwEpUD/sWfFWfD93hHlB2ZpNoRuQCiz2az6Hmwvb1dCc08xveScc8CeMiFknvIRPjqaEKqLvCSqtm3/Pg8RuDXJP24pJ+7+f/v2Od/LqX0NzUnBI9exwc4Cej/mGiH/Sgnk+pEigsuKGJ9fb1iQMbjsfb29sKbttttvXjxIppwEOOzCIX7wJLDHqOkTDbFQ5BQwHxSVCnNexx66TLKQp9CCpn4HLKp0+kEHEVJycdzD8+ULEIAPCeeHGSC0QP+4r18gRL3wxg7HAUN8FwoN4p5dnYWqIF5xiN7JaL3yXcUM51OI+fPmGDACFt4Vtp91WrzvQypF2DNwNnZmXZ3dyPWBwlyMG/E7qDIWm3epZhz6UvAWMBXsDDMZQRFZDUoXpyMlVQl/vjZY3hP9xHO5QVArjPODYGU/T3z401ThH9DcxJwM6X0maT/VHPl/5WU0k9K+kTSj96c/uuapwd/T/MU4Z993fXdk+VEBv8DfRhEBBxhYbAgYSg0AY6zZ+B0Oo3JYmUZCuPhBKk8fz5PxxRFER1tYK/JO/vGHY1GI8ILL0byeJN7UaTjqylRPEhIYnTWwiMgbOkFWnEPh9JIil2a8d7+bsBG76Pv+XKQB/NCrE0MD//icTPxO4aEA6HHw/szoLSgAMbci5U6nY52dna0s7MT4wCcJ81LPwFqKmDv2YjEWXnm5/j4WJ1OJ9aSsKAIrw0PhFKSrcHYuXxKZf2GF4XRso0xJaRBhpkv3pWYH71gLt0Jus7gkJinvFlLfrxpduBP3/KnP77g3ELST7/JdTkQcodoxItStdIPos1TXSgaws+iHmq8z87OKhb63r17evHiRQwsnoZ99PicCcMroIQnJydqtVra3t6OSQIZ0JbKY3sMDB4Pr0ehCjAYw4VxOD09DU/D9Xgu7umQFmEDoucQEIVnnAiNiE9d8UEIeJbxeByKiWGq1+vq9/tBpoGgQB2gHY/5SZdS7edkMHEx74r39LJtDC3QmBLntbW1WN7LVmRPnjzRxsZGfLfRaKjT6UTpd54mhR9g5ySMBwaTtnNwMJKiL4BUNnVFOZvNZvSORK7gWyAOSZ16LYFnUvifMIU5c2QEmnJEgVPl74zTouPOVAwiKG6ZLy8vg4RxQfcedVK5Rt69ye7ubsWTk1MvbvK4Djc578mTJ3r77be1v78fBUKk5PBCV1dXevjwYcD8Wq3cO5B7+Uo0vAhekxQR78NWXMT/1M/XarUIKTAs0+l8O3MvIMILMkZ4NC9tJZTyxVEIzmw2i3QlMBNvThjklXTwL9RVnJ6eBtkKHKV+wAlM0IQjPMaB+zqhS78ADDzcCggJ+E8BUL1ej/0Mgf4ffPCBlpaWIrty//79KMIiK4Lx4/2WlpbC0x8eHkbNCXsxoMCkL71AzWE4ssCuUrPZvG8EWQKKv0CO6ABjiyzj6ZkfxhWDz3jAo/i5jDMG67bjThgBHhjyjIfGM6BMrAZkPTfwFMuHsqGQkioEHH9n5x+QBnB8fX1dg8Egrnl9fR27xjDYFK1sbm5WGGd2CUKo8fh5P7uimK9ZB5qCXPB2Hi64R/WUKcrIO7ONGnDYBRUBxUMjuPANns2YzcryYQyRNI/h8ag8y2xW7nDj6SmU2IlTICz/UCaq7fCkq6ur0SqccApvTG0+EJxxpXa/1WqFghEHn52dqdPpaHNzs7JwDM/P9Tw0wTgSlhwdHYVy7+/v6/z8XJubm9rY2Ai00mrN96wgJGP8zs/P1el0KrUvFIEhL4SivBMl73m8z//O/sN1SVU07fwL4QbhwaLjThgBLJfH5Z7y8LwnwgW8QVBROPc4/EzMNh6PNZlM1Ol0tL29raurq1D6jY0NFUWhwWCg8Xgc8SEGoNks96a7vr7WRx99pHv37kWJLwbs4mK+TTcZDvLbvKPzGq5EXmYM7CYcQQmksmkJSojxwJC4R/VuPMT5ThzB8sNn+AInrgOyIHZ3zzeblbsdERLwfdqpU+zE6jqeBaWAkKRRCMYLNNFqtYKHcIMu6aW0H9embfz29nZ0cO73+xFa5WW+Tp6SNgSZYozZBm46nYahevDgQcwNnAVjT4kzCsl1Op1OZfGWIy14An7nPQl9OTAAzL2XoYOwJMX8gp5vO+6EEZDKSiuHfCiLVLYe8/3qibMhX5zcQvmXluZbMrGjMR6YLkPEn3jDVmu+/9/Tp0/18OHDWAx0eXkZW5LjPfBYRVHo8ePHYYggnCTFHgG+lJUKNZRyPB6rKIogoyDDINwIIYB6vCOKD/ylPx8ogRjSvbqkMAyEIKAIYDlK6my0GwM8qcedcC6MPd8dj8dRjYfRI2U6Ho/DSNLmi7g+pRThgFRdMuv7ITqCcgSyvr5eqc3f398PY+fhJPPnzocwlPkEHVAY1G63NZlM9PHHH+vBgwexZJxKUBavgbhINeLtr66uos+hh068C4VYTvphpHgfD5lxOJICOfG/k9G3HXfCCOC1ECpeFJiPgnD4zkLAH5TPCy0ajfmKMxSOaj/iMwgyqslarZZ6vZ6Wl5f1zjvv6NmzZxE+YNG73W4o1tHRUfT/73Q6kkpEghICu+v1euxCjMIBhbvdbqRxgIbLy8thqLwQx/PPoAg8HOMCE41HYiwJf6ju4zlJFboXxwi1Wq3o0kS44UjHC1yItT2jw5zisenIhCek/HY6nUaPxW63G/PO+zgZSC9IFIz+DQ6hHfJT+k2XaFLF3ANjjMHnGc/PzyP8o8CnVqtpOBxqNpvpwYMHGo1Gmk6narfbYcB9KTWVj/zuWRMMTx4GQjb6WgvmG8MOyoMXwrF5jQuGMc8W5cftFQTfw4NBcGIQA7C8vBxWM6X5KjvgnFt1SdED0Fl6Bgr4yyTgVbk3dQJAv1arpffff19FUUQ7ca6BV+t2uwH7jo6OAroz6cC9Fy9eVOAki2xWV1fDO0HuIEDunQkBeH4Y85WVlZeY/vPzc43H43hejCJrGTwH716U2gBSYFyXdBbPm+9UjEBjvLy4hRACdEcGRFIU0KC8zBPjs7u7Gyk+llQ7jL68nHdj5plrtZqePn2qo6MjbW5uVghashj0CTw4OIj59k7AjjpAFHhUiEmcCQiKeoWTk5OYO9+A9vT0VIeHhyE7bE7jZCThJsYGhOZVnYwzJekoP4QgsuGGGKMtVXeRzo87gQQkVTw/FXe9Xi8gFJ7NyzwlBYGDYHMd4CakkxMlxK94t/v374ciUs3GGoKNjY3YR3A8HmswGESlGAPM+vDDw8NYCMQkYuERJqw5SuhVXR4PAlWpWcCoeEETnhYSEuKU6kPCBS9AwXDyXMBrqvDcCKJweEc8FaEWKEJS5ZkwsnAIzeZ8rwUYfMIAwgMgMkKMgTs+Pg5j5HHy3t5ebBQL93B4eBipVowk49jr9dRqtTQcDlWvz8u5eSaUE2hN+OIGQSozUA7PgfhAeLoawc1gwFnwRYYBvgHCkPDLi+FwJMgoyu4LrXAMjB3Pm1IKktFDyNuOO2EEEGYErtvtSioLUigmwVPiZaQyLgX+8eIHBwdx3ZzMms1mkdpaX18PaE/mARacnXU++eSTSAP2+/2AcCAE/sbENJtNPX/+XGdnZ2q325VYz5dBeydb7kv8z3uR1XDB4XqU6HrjDe7BuMK1MFY8O/XwoIXxeBz5bNASHt1r24HgoCxvLuIlx+TB3RgQdnn8S1w8m82ikScKTEhEN2YUzmsVINsYdy8+YhzZE2JlZUVXV1c6ODgIwwfH8/DhwzDkFxfzLcc7nU4gPYwvLedAgfAkrEDc39/XvXv3tLGxEd6ZHhEgD7IyZKV4XmJ+aY5K1tbWdHR0FGED4yyV+2oyhhgEOBcMJNfFqC067oQRwBpjsREYyKI8jUWsBXnCgcdjYxE8CNaWst/V1VV1Op3KVlwoBMw08KvRaASMx7sOBgNdXV3FnoJ4MrynV6hh2PAo3gHIlQqLjyfN41UmE4IOhaJSbTqdRkyIx5JU2bLMU3Y8j5OvXu7Kd/C0jDPzAMvuRCBjSYHM1dVVrN7ke9QgIMAYD6o8ITqlaiNVuBXGmoIfPJynCukC7e+BkfjmN7+p9957L9Acy7ZRNki77e3tSP0VRRGLuFiazH6PpODIqPDc+/v7evvtt6NmgwwQhghS2T01KMBRD/cHTXIdSfH+GCGMFbwBsuJOc9FxZ4xAt9uNl2J7MBTS85/EavzsEG06nVZSYrmQYoWlsmwVz9lutyP+JT4lfUUTi729vVg8QtoF8gbCiOIhDBoC5DUHnl0gNJlMJuEtPH+NAKB4FNHAG+DlHD3wc34vjxnJqngxFjXweFu8NLB0ZWUlBJE5cW6FijrmxQ0gwusstQs9nl1S7LnIdfDa/AwLjyFBqTY3N3V4eBhp2M3NTe3v72t1dTWq8gj92GGauorBYKBaraaNjY0geXl24DhhJj/X63UNh0OdnJyEcaPQaDab6ejoKAyTIytHr7lc8y4YdPakxGh46OspQE8z+vJt5gF0uOi4E0YAi01+HlIPAQYdYNXx1HkdgZdfktaDlZ3NZrp//360mQL+AqdZlsqEUK1XFPM9ATc3NyOnC2zt9XpxDY+9vZU1JCWeHEFBUJk8PEmj0QjvhjWX5grz4sWLClJgdRwIgvjT40SyBrDOTjRKJQ/hBJ5UXUGXPxMCydh6uoprIeBwOJ5tcNTg6wlIMzJOjCvzcHJyor29vUAH9+/fjzkGrTHHZE7eeustXV5eRibnvffeq9T6Y1joGk1fCHgajBPQnefBGPM7Y8K14VFQ0F6vF7/j+Qm13IAj4yg7XALy7Glfxt9DNN8A1dcn/P8iRTgajWJy8TTAcfdoCDFeg79Np/OlqLCvnsMnHUMqCkOCYFMqSpNKINinn36qg4ODSkkvEBVv2Ww2Y2NRSZXeg14E46lOBMpXIWLdYd6J3WGoeTfiZ8ZJUoV4cjKPMfMWYfyNA+PK97gvBhXys91uV3ZtxliTQUAgQTeEXqyuAxJLZaUeRpX7QgI7hwCywMhAbF5dXUU7eA9ZKM31Kk+UGllaWVnRaDSqFD/BzDMmDsEpQOIcz4pQF3B0dKStra1AUEdHR+r1ehqNRlFJ6J2UQTDMC/97DYajWyeHCc28MA6j4ovRMOgeHi467oQRoEDF0xwosdf/IyxeIokFZQJ9/TrFNjDTeKR+vx+EI59RJyCVRCWxfavVipgQqEXNAMuAKVulMg9Skzjw9PS0skDIFdQLnjB0vqoN8shzvUBxqSx7hrF2fsO9NcIrlfwJKAnD5iQt1wBlUbgkqZJhwYPTNIPvYEy9DBZkg9JiNBgTQh0MHsuvWXDkKBHFZO0F1aB4Ugww5B5jzL4AGxsbMSakSb3KlHH2HaLcKNCAJqWkfr9fGbMHDx4EmUnBmhe+MR4+j17wgwPgd1aOgjQZY6+WZIwwIBgq5vS2404YAawiRAkGIM/r44UQXpTFFQe23beLBloBHSWFMF9cXKjdbsdEP3v2LMINhJ6OtqRyqFzEixDrA7mpLEQAvKDDWVos+2w2C+/IhLLOAZ4CpQXGwoPglaVqA09KZTEyZCUQLuA6x2w2U6fTCUHGKKMkxLVwKQ7zJYWngsRkXtywkZPPwxkMTFEUlVZwkgL5oXigh+vr+QIergtnAQ+0urqq8XgcK0bJ0sDi+8Ie+hg2GvPt1L3ik8pCiqR4Hth3nhuZdI8Nn8W1MGygUC9qk/SSwwOdMV/IGkQgoS8Gwc/30AIu47bjThgBqbqtNtYUr+FezJfTMvicgwB6R2EYX68hYDJh7+kdf3x8HPfGwvM8TCQEFwQQAoEHoVjIlYZYzash+T7vfnl5GUgB3gG4SiktsJUiGM5FoXhOmHSEy9cruEfwtKoXIzGmQEoMgnMwTkIy/u7d8OwYc4STczAOzDlpMBqDUPOAMJM2hs9IKenw8DCqAKmPAP08ffpU/X4/5uPs7ExXV1exh6Bnihxhttvt8KTMmxtMnAOGkXFjpSN/A91KpWcnperhDxkw/rlMM1ZwCK7kbqi9mYvPHRwG8nzbcWeMAMKUp5yI5RAgz6diJLyk062ge1XP4bsCra+vR3MQjAf5XwYQ9hrrDjHk9eZANd6h1ZpvPjIejytdZBBgLLdUKin3giBtNpuVRTJU8wGB8UrOrrvXl8qtxjAACDuIhufx5cQYEp4ZhSUd6kVPKDrpMsbfG6k0Go0KCnAEwnzynL7YxyvtgNOEX1K5AWitVgsExvuA0obDoTY2NmKJMTzCbDaLrcYxRow3MJueDrQc94pOQsOzs7MwNGtra7FHoYehhEX+PI76HB1xuOJzQJZiXM7OzuJcN66c54jlVcdry4ZTSr+QUtpLKf2/9tl/kVL6Vkrpt1NK/1NKqXvz+XsppUlK6R/d/PvvXnd9DqygVCqFpMjB+mDm3/PqM7wOTDdwrl6vx+aR1HNDoCwtLcVOQ17Cu7a2Jqls79RoNPTw4cM4B+MymUwingSOOoxGgaVy5dpkMgkF4p939KGQaHV1NToMY5SImZeWqrsau7dASXluLz7CWIJYMIx4ZRTCOReHnXhI/wyj44aWWJ35IDziPhhQIDLH2tqaNjY24jOvsuQ5cQJ5+o50LHs5eD6ffR/w2o3GvPQX0jWl+YpKsjmQmV6+i2Jj3NrttpaXl7W5uSlJ2trairCKIiZX9np9vnEpaMSXOLth4D19DjCurdZ8cxOIRuQVpIkTZX6c41h0vMnagV+U9Ceyz35D0g8XRfHPSPpdST9rf/uoKIov3/z7qTe4fuVhyW/yuS8iQskgmlAiBox1Ax5DURCC95YUaT0G3qHf+fm5Dg4OXiIOgW94UNY0EHZ45oEQBEHxeJxnZP15t9uNBUMoscd3GDJ4B54HBtxTjyi9e1aMIZAWL4jhId4lnHLC0q/vpbCENBhqDIEbEe+X4EgOTy0p5sfDBKls0761tVXxfMw3aI3Vk3hkFJkxgDhsNptRAYihJRPEdweDQawYBBGCFAm/mJ/BYKCjo6N4TghlyDmU0g0thV5kEFiV6suPSami/FzT07kYIpwX9/GSe6lEx2+CBl4bDhQLNh4piuJ/tV+/JunfeN11XndgyaTyxd2jQq6h8EySn0NGgJf3GgEQQr0+7zLMBFLU0e12K6vhUGzgFRNAGSueAa9NuozJcIII6E66jJ8lVXLPkJlYcV9bgOIQswJXWXnoRUIoLQojlQrEOSgjvAgLVxh7TzNhuMi/o7zudSAvqTe4jZBiDvDmoAzmnxjbkQNjyuIoxk0qU5x4dScACdF4L2oJeA8KnzxV52Eo48q4eaoNpATsRwadqGZuuQbhI/dCxvndZZ93w1FxTc8OkLnytCVOCgfUaDRe2VBE+mI4gX9H8z0JOd5PKX1d0rGkv1gUxd9b9KVk+w7Q+BGBd/YaBcj/d+ba2VGUlcFDmIqiCMtLTIsgUgnnXY2ofnNWf21tTYPBIHb/QXnv3bsXMTUCzGRjWAhrMFYQQkD6o6MjSWXfBGA/CIEiI+6H18G42biGYeTZEQyphI5Sud+gp5R8nB0Gg0xQYlKoGDypTDtyLZ8Xrsc9nQNyngJh95COYi7vwQDPQpkw70U9SKvVii5KNDehngKyD64CQ46hQf4IIz2cYn0JfAJzikemZBmugGs5WnAZBu2Bet0oeNbBxxHk2mw2X1pzggxSV+Hrbm47PpcRSCn9J5Kmkv76zUcvJL1bFMVhSulHJP3tlNIPFUVxnH+3sH0H3nnnnQKv4y/tNfN5OgsSxVNJDCyxkcedEDzkeWHGfdGGs8QIO+kpUAg5fxQfxtdTZ5LCY8F8M4HE8yjp1dVVlKx6NxoOFIFnwOMilF4WimLhTTAyrnBOxOXIAHSA0LgyO/sMIlpfX49+/NwDdIX3JlxiLB3a8hnC7t4LRd7Y2IimI4RGwHiMa61Wi8VUvV4veIHT01NtbGyEsrLkl+dgDlIqN0ch5ub56vV6VICyVJ3x5hl4H8YOrshJ7evr6woihMvCkXhxFwaXwwlaZAMjTFqa5eM8j+uKX2PR8R33E0gp/YTmOxX/W8XNkxVFcVEUxeHNz/9A0keSvvS6a3nqI/c6UrV89ube8b8PCoPuENEXabA4JU+XIYjkcTEa7rFOT081GAwihsaD1Wplt1ivRaAc1Rf6sMMR3oV7shqR78E5wEEsLS0t7B3AO+MRPM3Kszux5FxBbhjgCL6vS1YAACAASURBVDy8cjIJIo7vc02eCeEmTOEaICMnD/GykirP6rwBP3sLd9KAPLd3luJveZ0JXhivDJLCsG5sbETFJ7wP4+Nkp6RAGcwT9QwYLTwwhsTjcv/fjbRUthmXSqX1cfE58awNBrEoimgM4yl2DxNz4+DHd4QEUkp/QtJ/JOlfLIrizD6/L2lQFMV1SukPaL4z8T990+vi/Z3w42CiOXLPxfc9TsUqw8hjbb1lFIaCa7L5Z0rl5qBMMsaIn2u1WhSaeJEM18YjeONSKu1ms1llKa+PAROIYkvVXooIDOMFLPdJdyPn38dDu6dyA4Hy++o3zqU81xnym3mXVG3KAS9QFNWKQXL/eQmsk6DMFYq+srKiXq8XiMAVgrAKXoJMTb/fD1TCOJM757vcG4XGeDgRKCl+Hw6HlQIvuCFP7zJvGB/GjzlA+ZFTDAfz6fPl57tR8rCBVCzZEM8MuGH1UvH8eK0RSIs3HvlZSUuSfuPmBb92kwn4o5L+85TSlaSZpJ8qimLwunv4S4MAEFqHW06eOPxlcvk71/BCISAgeXcOarMRIg682NXVVcTrpJucSUdA8NggAxQTqCcpvAXwEAIHFOHKyTsz8TYflZAJgUIo8vSb/82NpROQeDaP2yXFu2JsEVrnO/D+GAy/B5CauNSFnPH1bEGePQA2s50cBiInH9kgtijmawdY79/v93V1daXd3d0K0skREnPj9QfOASwtLcViMUhAPDBZAvpScB8vAiLUc+/P/DL+uSFwY+5zzPUwRpICeVArgVw5gnjV8SbZgUUbj/y1W879VUm/+rpr5odbLocvDuWZGKksj/UllXh7XpqUEek1UAEChPKyEgy4yrJeX0jEJKAAtDuTyrpuV7irq6tQIDwKz8LvKBpML0pESun8/LySknT4z3fxxM6deI28e8E8Xy0pQhaMJp7YY1D/nAIZ527ccDE+3BuD5AQh1/Lwz+/nJCGoBIN+cXGhBw8e6MWLF4G+8PCEW81mM9Ya7O3taW1tTWtra0HU0b/QPTj3yjM7GAFqMkAanob15jOUoGNMWUDlht2VchHh58jQiT7/3f9xPl2SkRkPt/Kio/y4ExWDvBCHwxiEy2GjVFpR1pezwAZlAgW4AYFth5RxksbZWgYSWNftdoNsIp5MN2m1lFKwz0BRYn+uyQIfRzBMOAqMN3ArP5vNAlI6x8F9MQZSuSYf4fYCJyfjPJvg8SdjxPkoLKlZv17O4rviO9/Az3i5vOMPz+Uow+eYZ5MUMB0jDFmJIuDJSQdOJhOdnJyo2+3GfhLvvvtuFFdhmDFQjLujE4/3KTWH2MXwYzwchp+fn6vdbodB5zklVUqzPQz1ecH44oAcATuCIpyVFM1OqCcB3b1u3YB0R4yA9LIhYHKBtVLVALhg5ylCOv7y2cnJSSWPjTdjklEASEMIQieniGsbjUa0CCdlBRzEI5GGwoARVpBK8ziRdFdKKSAmq/E87vNNVtwAkH0A8ZDmxOtz5HwBcSLnuCDzTh6eoaiMl6QKdHeIjSH1OUDpc/TgGQcUj+thMJeWlmIXKDZu2dzcjO7E0+lU+/v7UaG3vr6unZ2dmL9PPvlEOzs7Wl1d1fvvv19pBIvMkEpGeZxvYkx9viliwihjgAgDWOhD4RH38LnJx5XxQfZxfnk1IO/lDoW0ITKAkXDEddtxJ4yATwYD4PGQewyPx92KE+8B36gph6hj/0CURSrZXt9yjAorlEBSTCYVaaT/qBdgWzO+h1CgzEwoqMBzuF6554QlXoI0GMLBdRA46t3xxiwk8hy0E6X+XighJBaC6MYA4XKm3BGGx6gcfj+UCoPm8BgvBVrjfv5s/NzpdOK8g4ODyBhsbm7qyZMn4fFRMAwjToEQsCiKCjdDJymvUkUxyUQQjsxmM21ubkbFHrUjpJ4Za1aQ8n6UFvNsruygL8YEGWVc+d3JValcyIYeMNa+dgF0/KqGItIdMQI+8Cius9m+TFUq4ZJUzYNiORk8FzIm0gtlvHQYa+lNLL2KsSiKSh26pwCluXKwMtA7FEHSeUzOmnj2ouMam5ubLzVS9Zw2xhJB4vm4vjcUdRLKDUhOZnolJYaUEMANss8V4+Pxrce9LuB4LEd6XJN5zonOPGvBuZR7F8V8jQCczdbWVsBwimfYY4LsztXVfKt5thWj5NYJObw1xJ6v60emSC1iCIuiiBWOzDU5ewwMqJAKP4ybcwMoMWPoRpHPmXscoztNrknj1H6/H01ZfG4WHXfCCEgl1Odl8awOG72SzuOenGihom40GsW5GAW8LMrgVpfziDeZMJ9oX2XW6/U0HA4jdcaksHCGiSJ1BHSlnz4eRVI0MUWZHeUwPq48zhfQ24Buwc5tcGAkuZYTUR4qOF8hKTIbCCVGzpt8cF0E0TMaZEtQJBSSueO6GCb+5k6BuWZet7a2wkgdHx/r/v37SilpZ2cnvCAcBgrLRq50Rzo+ntevtdvt2C+QMafegIpJYnDkBi9LW3TKwTmfIipPa7OOgfb1yKy/I8VHyCRG3zkyRw+Eh8wf3wGJes9HDw3z484YAffkeDYEkYHwnm05c+qFEwhhr9eTVPayI0XoBM50Og0IjWeo18sFMJKi9TOC489D6oeiFofzs9ksWGNJEdPi5VAQZ3BJhV1cXERPfzw6XtA9J2QhhoYClVxYMKKSKpkInsWbtbjwcW9fwYiB8cIZqQwVeGavr3A0gbGdzWZBaGHsMDp4PeJg3pEdimjvzUpPz8SwbwDvv7S0FL0FfIkzSLFer1eKrS4vLzUYDCI1yLUYF1+hyvmEZ8hut9uN52C8HDmCGhgvDEGevnZE58bax8zTwY6U1tbWdP/+/VgcddtxJ4yAxzWSYtKZIIdBTh4h2DSg4BpuKCRFm2dKS51w5PeTkxM9ePAgyoolRfyHx4YAJCvBFlWQSkwiLDNGhV72xKIghaurq+g3MB6P9fjx43gvPALEpBff+Od4KMYJRph8P9DRY+98XCVVUALCDpKA0EopRaaD/702gBJqZ9UZE+bNYTQIwIlSNz4Oez1MQBF5F7YQ57PRaKTnz5+rVps3I7l//76azWZUfLZa870tKBBi7QFkJkZBKlvfSdVFbswBuzLlHEZK82XJrFOgQxRG3K/p2QEffx+PfG7cEHjIC7GN8WeB2KuOO2EEXCi9dBOlR6GdFed7lGli6bDarNjyPu+OMPCAcACdTqdSKJNS0vPnzyVJo9EoPBue2qEduWWe2RduEM+DYvJOxxCNCKNnARBGCC2fcMgsBA6iyIUJwUBZnPD0cZBU2ciCEloEPme08WBOXjp05hwQSC6EGA1Hfoy55+qZH19J6OlZb3JC63DWNHC9TqejdrsdBV9uSJrNec8IlgOD4u7duxebjvKcjoK4drPZDCXDePhCstlsFunm6+vrOAfj4ePE2CNbbrgZZz/P+TOcjqeap9OphsOhpHnIA5m96LgTRkAqW1k5m40Q81muwMBnyBCsqheqMIEIqhOGeGUvtqHTUErzHX5Ybkp4wM/Eeq1WuR8iBgzPwLMSTpycnGh/fz8Y6q2trfBIHHh4ahrwHlIJo33vAy8sca5EUvzOuHraj7EiMwDyIqPB9xE2Xw+Q56sdYVBQxD147jytC9JBKZEBntHDGedquIaTir1eL97DV4hSE4JTkBR9HHyBE1khX+DEPPAMVJoiU2QPSP/iZCABPUORUor+D4yHLzZj/uChPCRyPoXD5xhDkYdryMx4PNbm5mY0PVl03Akj4EQVk4aSe82zF6sgfAg/E4HgeTkrYYKv/GKfeNYK9Hq9IO1qtVpsHfXxxx+r2WxGOpD0YkopOAD32nhkBG1/fz8gIc8PN8DEISCsTOR6CAAKDMrwmJprelEN8SfeGYXBw+Q9Bhgvb27CeKGwvtQaIYbMk8osjrP+HHhEBNObtTDPPA/z6oSZVDb8dGR4fX0d9RkYJ9qfT6dTPX78OPZqaDQasciGhqR7e3taXl4OYpC0MmPJsy1CZOTj2TAVRNLpdGL1omcFOJBpitUIibybtPMqGEQcF57fEbGjE8+y0OVYUuUZ8uNOGAGEwA0B0MiVFyTgxBhsunsPL/pAAMjjorTsRUjRx2effRbGgmW7L168UKPRiGIjQgKUy4lAjMLJyYkePnyo8Xiss7OzSGNBIKaU9OTJk2j0ABPd7/cr9e00xPAaAd6JuBZlRoEZJ66NkUC5PGMhKYhBH/vr6+uoiXAyEaRDCIPBo/YBiM1nzl77NTBEQFeHr3hiiMpF1XXIy2w2i2dwROPhzunpqb70pS/pG9/4ht57773oAo1M4FBAf5SKw7OgqIxxo9HQYDAIL08oiQGHkENmQCB0QWKcx+NxICaMp4dWPsegMfgKZB9ZJEsDena0wRzu7e2p3+/fqn93wghIqsTQUgllPeUFeeVxsHeYJXZFSb0sFEPRbrcDGo7H46hFR3jYmBQWl1pwvPlsNt8gpNPpVPreO6ogU4DiEj48ffpUzWZT77//fvAGy8vLOjw81CeffBLcANDNmWJSnwgBSu9eE8EFllL16LAa5eOaIBeQE4JFxRvKRshCgZJnc2D4JUWmxQktvJiHdIyxvwPPB9z2jIeTmMiF8xYoDHG3F/+89957YdA7nU5kaD744IPIxNDyTaoiU66Lx+c5yYzAW6ysrGhtbU3j8Vj9fj94H/gVDBYIUFKsOwD2M/4oM6iVeXVk6/yJr4NB3p2A9TT6ouNOGAFYZ0kBO50MlMr6AT4jhuNvDBKLKIjpiPew5Lu7u5IUnpTFKJBPGxsbqtfr+vDDD6Ma0Pcw5HvEiAgmRgEPhuAMh8N41ul0qtXV1agtv7y8jD0Iu91uwFIUHG/q/AfwknSoLzhyhp1yZC+IQYD9H/CUZ/YNLoDcIAtgcV5vgBBeXFzo9PQ0kARFUcwDno/nplAKLsEVl3GAnOMdkRGMlS8wYq0AoYNU7ofw/d///Xr+/LmOjo6CJJxOp9HZ2He/hvdptVqRGYI/IqSAyB2NRrEDMWndoijU7/c1m833k2A8jo6Ogmcge8T3qIhEcT1sYO68LsB1gfECJbvy8/ne3t6t+ncnjIBUFSggojcB4fCXxAOh/Hg0Lw1GuIkl6Qc4m830/PnzsP4OP53Vp/7AYztKddvttqQ5iqGGAMU7PT0ND0HP+5SShsOhOp1OLIJZWVnRcDjUbDaL3K6kStyHcEEkORJwttnzxhCLUtl7wYtKgPoYCOJ5DEheqXl+fh4EnMfzeC3CKPfSzuF4GgtSEQX3NKCTuGRU/J0JNbwugJ4DvJez6FwPAzedTrWxsREVgaAHR05FUWhvby+McKvVCrL28PBQ7XY7VpJSoNVsNjUajSJT4ZwKGRcgOmHG1dVVVJkSWjA2XinpYwfycXKQMUE2nXNBd15VOnwnjIATQzw8E8zEMUCeQ4YQ9IU1Ti65UtRqtcry4N3d3diyql6vR69/wg88IJCXlYEsECI/zAaY7AwE9GYV2cHBQXiEdruttbU1PXjwIOAbsaOkmHhn3qUyHiVuJp7Eg7j3IGyQSvREjM3vjIfXo0NI1mq1yLYw1hhZ0q4YDJ4RAWbcJAUCQQn9/UAYvJNUlnH79uOcS02C13f4d4n/GS+KcyRFNSdrFkBrxPYplRu1uOLAraBsedPZwWAQ74ixYOMUZJB7kwlgo1LfqyBPtTJGhG1eR5ETqV6whMHwJriERcjFbcebNBX5Bc3biO0VRfHDN5/9Z5L+XUn7N6f9haIofv3mbz8r6SclXUv694ui+F9edw+pjBuZFCbTd3h1ospZ6lqtVtkzIE+pUGnGpH366ae6uLjQ2dlZEHLdble7u7sqivlWZCgZBUEMKCRYp9MJ605qcDweBx/ApHY6HT18+FCXl5fR1ZjuPJICHiKkvo02nYkxdnhueAesu08y3AcKwnPgtX1MIPgYf+7vhCfvQvqKg6wGCg55CsQFrmNIQHQ8R27wURqgPnMpqbLJCxWfLgO8z3Q6jS29MWbn5+eVnhEoHVuds5Grx91eWemEKgbw/PxcDx8+jI1Cx+NxGEE6TRGqUi3qcB0jury8XFmvASHsxtdJXVCLc2Kc4zLg8k8o4vqQH2+CBH5R0l+R9MvZ5/91URT/pX+QUvpBST8m6YckPZL0d1NKXyqK4lqvOfL12XgRJgR4iUX21A2eHwg7m82CIGJyEKyjo6Owqpubm1HAgwcnQwAEXFtb04sXL6I5BTEntf5LS0sR+7L7LhNfq9XU6/V0cnJSKWkmFXh8fKzRaCRJwQdMJpNgcknxeFUZXgcDhVDBaEMcgQYYO8YrT8GBhI6Pj9Xv92PciI+pckTJMBrkyJmjpaVyZ2fif8jIoigqq+w8POBZ3NtxLnwBBTlevONcBJwAaIXx99jaFQRjRcEO9f9eC4KSYizp3nN1dRXzvL+/r62trQg3l5eXA/GxqQmcA0R1s9kMshnjBMPvGShPl4MeMCxcDzSGgjtH5aiWa9x2fEf7Drzi+Kqkv1kUxYWkJyml35P0z0n6+6/7IvlSIA8xuMd1eCrgObGTp3Q8y4D3QUHJAqBIlAjjXfFQtJQGSm1ubmp1dVWtVit2rmk05gtFDg4OYt08G2TCpG9sbMR98HLD4VDNZlM7Ozuq1+e7In3rW9+q1Bkg+Bg8Hx8QEdVyvKM/uxN5XgsAz+LEEsYWAQfOt9vt4CVAJ/zvadnr6+uo0huNRhFqoPwYJAQSuCupYoggyLi2ZxJAg/ACnt2QFJuMQkqi2MvLy1pfX9fh4aHq9XqgBJAcq/0IPzDUIICVlRWtr6/H96T5ehR6CyALjmpOT0+jpoCVfBhSSRHuMB+12rzdOePjTg7j6CEAc4cBRl/4WdJLtTHI523H5+EE/lxK6c9I+n8k/fmiKIaS3tJ8MxKOz24+e+lI2b4DHORcgZC8FPGg99ondGCwYJPxXKRvyNt6NWK3243BPDs708HBQSg8zwN6gKwDoTDYL168CIKJe2O8WNLqPAIr2X7nd34nrs2aguFwGPljaZ6K6/V6Abc9O4CBA7rmjUbJjqDECATGFGUC8tO0w+sqUEIUrl6vBxeCoeO7xN++65BDV1/UBNqSqluq87sz3Zy7uroa6VgnDqVyzQP1HswpJN9oNNL6+nqFeDs9PY3sR6023yvA+QdQ13g8riwy29jYUFEU2tnZ0XA41Ntvvx2Ga2VlJRQO2SW9CqnoKAWSE/4FmeI5UHaH+fwNQ0tI4AgAWcQZTqflhqW3Hd9py/G/KukDSV+W9ELSX/52L1AUxc8XRfGHiqL4Q+wM48UwTvwB4aRyEQdWM19ZSIwklZD16OgorLmz4rDfrP5iW+xer6eVlZWI31lJBh9Qq9WCD4DlzXsIjkYjHRwcxOIWPBSwGmEsirKjbqfTiXZpQGhIQUmRbwZ+4809Bs2VH6jqSsPY8f2rq6vYl4HFNJyDMEsKtAC5hrFGKYGxjAPpSzIOZHDckzMXoBQQhlS20/LKOYwnZB98EOjH+RBqOKTSO6KkvV5PRVEEj4MB531RquPj4ygyOjw8lDR3Wh988IEkBYrgnUgtEiZSeFYURfBWyCYOwLMyGDr3/MgZv3va3Ily5h4ExXWQt9uO7wgJFEWxy88ppf9e0v988+szSe/YqW/ffPa664WQ8oIoNQfC7GwxKUQgHQQScJ/JQZgp/8VjAr1Zj390dKStra0g6Yh3Dw4OwgjwXQozWq2W3nrrrWgSAlTHC7FvHTH/eDyOf5KiJdbR0VHkvE9OTmLpKxOLMcTTUx5LSOPdc3wNBUYTT0CIBeGE0uK18/4AGGfCBrxUnopj8RPXwIPl6KLRaAQkJpTjHn5fnpvQAN4FZOcygSz4qkUIPA858Mw8G0VQpAfxmMgdTWRwOuT6USyWJR8dHYWhRE4gDTHQjH+z2Qzy0NunE+Z6ytVLiz2EkhTOB+OAkmOonVNgHm47viMkkFJ6aL/+KUnsWPxrkn4spbSUUnpf830H/u/XXQ/hxksjZLwwcNLDA2IttghHUCDBlpeXo5JPKusFYLsPDw/DAwKhtre3w9NB1gB5ea6dnZ1o1IAxwFuzuehgMNDz58+DNcZaUz5KSNHv92PF2/r6upaXlwOF8N4IBNfAsMEFoCSrq6thtLyuAuXCE7si01UZgeU98qIrz3agRFybZdwIPMoGL4LwYqS9rVqtVouUGsur+RwviQGEcIQbgfz1VKa/O/PuCIi4HQ9N2ObFWFJZqXl9fa1+v6+1tTX1er14LsYUlOLLkpFnsj9SWUXJ7keEGU5CskWeVBKaEH+Mg88nqUrnJDBWEMSOMD5vinDRvgN/LKX0ZUmFpI8l/Xs3D/+PU0q/Iukbmm9P9tNvkhlwK4oQYdFhVHkx4D+TiWfAI2AAms1m7CWYexJvISbNSTBywl4CDJwldqPQw+voV1ZWNBgMKt7OmfWiKNTr9QLGS3OFH4/HsZBoaWlJh4eH2tjYCIUmRvdUkL/rbDYL8g7m3MtoGSPPDBCrYpjc2DKuXtGHh6UIhpiftCYFOJQ7O0yF23BOwzNAfEYWw/PbGBtPc3olKT38uBfpRUeJXqQE54FxYq4ePnyoRqMRlZ3j8TjQx8rKikajUWVreKD68+fPgwT2mg26HO3v72t7ezvQJMueJcU8tVqtIB0heHPZd+IbRwBCYrydJPQskhsGn5tFxxe678DN+X9J0l963XUXfC9q6SWFYKKskD1ObkmqQC7iSuJp2HbSgldXV9EbAPgHhO33+5pMJhqNRjEhT5480fr6elSYIaD1+nxnY0gnBBwks7y8rGfPnmk0GmltbS3u0Ww2tb29XalvZ4IhKr1e3tObFxcX6vV6wZQzJnhHyCnQg1cAeoYB5cWQoETE4ngf9yAsLwaB+E7IR0dHWltbC89EHM8/X1bsHi7fgZm/Ux6MYIPEICT53fPuTrJCtAGLORqNRpREQ5QWN3n5ra2tWBvgRgIFOzg4COXd2NjQ2tqaTk5OdP/+/agz6PV6FSM9mUzC8BAqMSeEUswxIRNjDLrhXOA87ytVF3wRguRhn8+TG9T8uBMVg1K1uywEGlbP4yKMgOf1iXFRIC8swXM5PKR01Cva6D2H9bx3714sLoK4o5U4zDKGhWXIKc17EAyHw6g5J88/Go1CONmFGO5iNptFOMCzAwHxaEBRhDhn2BECxhLFcpaZZ8a4AEchGjEqLM/1slmMAwrHsyGYhBpUd0plr0HG3g0415RKqO5Czfe9EpG5duIOB0CaEO8JHJYU4+npYMq6i2Je599qtfTs2bMgeCVFSORKJimM8XA4jDUgcC0o4Wg00nA41KNHj2LuyCRASrvsI/+e/+d3l12MqqfSOYfPIIudNPRVmPlxZ4wAD0tKAwvI33ghBpFSUYSAgXHSDIH3eAwvWRRFpJ0YJLw9LcuBh8B6dnlxD3Z5eanDw0MtLS3p+Pg4UoJ4KLfYOzs76nQ6kVZEwYmVpbIQxpl3Z3t5d5APRs7r1B0KSuV22FJZQ86zcQCxQSGcg9HEqHiaivJX5sCVk/nCmxOiMZ+5Z/LUoBt/X2iDAUHwuQccBoIPIuL719dl9+RGoxEEMmNJ+AY5W6/Xo6U5ctXtdjUajXR5eamDg4NIk0IESwojTZh2dXWl4XAYvAPhI1yRQ34vhiMFjFJjIJkTxsrnF5mHIPV5RB5uO+6MEfAXYwWaew0Eyi2le38IN/cuTojwd9aOU2GGFSYD0Gg0YpMJhH9jYyPiYuJYsg94Fbwxz+QtsfiM90SoDw8P1el0ogpxMBgEqw0qwJsjGPAi1Lt7hSOHK6uPLc+FoXCUwNj795zoIowAmhLecB7zgMB6FSe1HFK1XTzXhmfheg7FMXzUQ0il0KNYIDJ+d8+KkhFOch+87NLSkkajker1eszz4eGhZrNZLDum0Ij3pHy81+uFI2FF42Qy0f7+fswH7+upOlLRq6ur4cDyzAiHV076HPA7hm5R2pDy9FcZAOlzbE3+RR/uYYjtF8WyeAi3clhJvDYQl8HHsjNJw+FQOzs7Go1GmkwmlfOJOfne5uamDg8PI7RoNBrRgAJCCAiMEsE1IHRMBBVmGAcWRw2Hw3hvhJ735B2p5+d9EWCEnv9RZjwRCskYebEThoxrAo+93oGMwdVVuXGqk0/wNh62SGU2hnlxA4PRwmjA1HtWw5l8DCfsP+/GtXh/xhtEQIzuBU2OAFCOtbU1ra+va319XZubm3r77bfV7XaVUgrDTz3CdDqNDkT0Ezw7O9Px8XHIDE5CUoR9GBDWMrBNGuPEHCCzhG8uF15f4SGBpMo8Y5BB1d48ZtFxZ4wAE+zVULwgMAdryeQBc1148Cpeqon1dLLRyazRaFSJgd2y+z329vaCE4AZhpGmNwCVdOPxONYMpJSiXpw+gzzbyclJFLHgZViVSAyJoOF1vEgJtESa1OG4IwEfC8/GQFp6CtZjYGeXgd2cA8EqKTIPKDbKSVyP4iHsDnNRcM6VqisqOZ+QiEIc3s9rShyFECL4yj6ei/0JKOxptVra2NjQ5uZmpVGpNN9rAsPPvWgyIynk5OjoKIhlrt1sNiN7MplMIlwklKQwzJ8v5308/IHXcKcIymMeQEQeRn7hdQJf9AGRgZI68+0xEMJFbprcMddA0L2eQFLE/75xCH/nHzUH/EwziHp9vokJnpy0GIKDV5JK5XFvxCKkWq2mvb09DQaDeO7pdKrd3d1oSQ4yAWWgyJ4BgISjxoH34+8O9d2AeWiUhyfO3ruXptrO0Qe/eybGWWxf0CJVDTXX5/v+HUcOkirQ1wtjkAecBtdzpONrJDiH1Z1cl3u1Wq0IvYD4tVpNg8FAe3t7L5Ws12o1vfvuu2o2m7E3wcrKivr9vmq1+XZpBwcHUbNCIxLCTEILUNfZ2VnMIw4wT/N6mEvBGmNEyOWcgcu+O4XbjjvDCbgAu+dCcJ0ZRik8/YXxQLA418kWoiUXnwAAIABJREFUz8kSTpycnETq6PDwMGrGWQkGDEbwQBhcg3p/CCInYabTqTY3N9XpdCTNS0rZNxBFJgVJSSn1EIQEGCk8GwaRsfI4mO9IeglCeuYDRXDUxXODDHh+V2BHFn5NvwZhEXBUqu6y4wtf3AjwLnhqZMKzQv6cXMvJY0ILiDHe0Y0fCu+EMkab37vdbjgDv76X4bLohz0jOp1OzA/jTCqTsNFL35kbjBmKzPcxum48mRPGnrlwhEM4SQrWjfRtx51AApLC83sc7Arl8aRbfx8oGFI//+LiImCaezoElsothAHoNhwOlVKK3W0xBngTMg55lR5EZqvVii5Gbpi8Mo9n6ff7sTQZI8e6eTIg19fXsbUVKTCEnvvmBTvAVP55NyKel/FkxZ2vp3eFdAOQ1wJ4ahJl9NAkV1RHN9LLYZ1ndjDmrBdwR+EK/qpn9TULGDbmAIPuPSyWl5fV7XaD/6HPQKfTibCRilQQlPeeAPJTjYlXZiUq7+gkJbIHUvWQAKMJguI5fQzRH/7GuGF0+dui404gAQQDQcajcnjNAIInqZJm8fXTTC7xIPlb96aw8M7i12o1HR4eRusvjM76+nrkwQkDiPdheFF89wIIxPHxcWQEQCasXMNToJh0M/KVb0wo3or17CgEguTEoMeYKJRUIipHVSAZjLAjLVCJVCoXQulC53PopKxUNudwFOLogNAH0k/SSwbBkSLoJ+cskAU3HMiFIyFn0z3bAQri783mvNkMtR4o2mQyUbfbraSYKZrybkTIInN3eHgYoQrv6XUYhKu+N6ITh+7ZOTB8bhDQDRyFI+tFx50wApIqRoAByP/OQPjh5yIoxJjO6CP8rDXwXoPAwMFgEJPsA355eRnLgKlMA6LCXlNKW6uV/QRyxfNmFyml8BSffPJJtAkj7sWIEbNKZfqHd/JafwSd+JXxcl6BA8UAwdx2nhczobzuiZkXfycn6fi7ZyQWzaMrP+/Pz/yN8XbiD6+JsVwkM/zvZbbuRd1QcC6hnfNAyBqkIDUCrVYrNqb1TsKc5wQm1YUYZ0+D8m7AeEJIrwplbJEPR33oATLInGGAvXoyP+6EEUAwpFIggHJ5XJfnSWHWvazSBde9IspO6Svwm+YSpAq9EgzBgan3/ndSuZQUqO9CDwqhISmVZbQ6I6bc39/X7u6uGo1GcAO7u7txDxYm4TFo+ukxKB7H6wUYT8YtV04UCiH30l7/fv5ZDrvzucwNhP+PAcAj+308C+Er5PgOBpl7cB0vrHGD5qELzoXxcPnIOQjI3el0WlnsBGpkvQckcbPZDI5geXlZT58+1XA4rHAAnkHy8AwZ8pASGfc5xVCwNDlHP7mhwsiDMl+VIrwTRkAq01geRzrkdMV3q+hLNBGsPD7iurSnwnvu7u5GEQhkzerqaqwvYJkvCoindkNElsLZfKAjk4OQ+r6IhBYe4qA8CC2FQWdnZ4E0PK+PYLlyOVGEIiDoGFPOdePKwTV4DldgV0pHBZzL/w6/c6Pg5zkX4yQlz8H/QHSEmueG05HKgiV+djINDsD/5iFOSmVzDklhBLg2VX4oO23pmdN6vR5IUZovEnrx4oXW19cjDPW1EoSRjCHGBvmTFKEqiA/Z9fd2HggjQjiSO9VXHXfGCDBhOUTFinm1mlS2qUaB8ZDAJAb8+vo6PC/n1GrzXWAGg0HE+7C76+vrsZyX1B6LhPASxFrb29uVwqF2u63hcBj3kBTQsdlsxi5EVKL56rZ+vx/CRRoIMgxDwCIUDAwC5hDcURW/OzPsMJzxcWbZK//8PP53hfaMjX/HOQM3Rs4pMNd5diIPPbg+z+4KyzVRBEg/ruNoESPgzU9zJMLzMcc0G4Ew9V2opXIdymQyiUVOs9lMW1tblXUKpJ+dh/GmszgoSrBBnNSzMFY4BhAnY4aRyI0u88o1bzvuhBFwQhC2dpFX4GdPLUnlBo/SyzXrjhqAmIeHhyqKQltbW1pfX1ej0YjlonTQ9QU8EH6eVeBvKc0LgbgPkJOmEZeX873uvSjl/v37kS6kcQWGiFWCPCtbmo1GI11fX+vBgwexUo6ORU6ySeUmIwi6p5N87YMby5xwyueHc/jZlZYqSz8HQ8Cc8FyL7umpTa9F4MDQu4EoiiJqSzxTQJUgcXNOsOVIhHfIMyuQpXSf4j7eMNa9M2HD9fV1tJZHJsnqNJtNHRwcxOcoOj0Jnc/KMyvwM4wHY+QeHxn0FajIQY74/LgTRsAFggF3pQXiOmE0m5VNJfjZc9BcVyqJOZphNpvNsNbAOqCXezNJ0R9OmhN9lIzmsBKoT9XZdDpva/bs2bOYbGLK2Wy+0cje3p6ePn2qs7MzPXr0qNIyi+q86+ty000yC6w14B09PPGwwJXfWeQcLjqsdGIpRxVcy1Eac+OxKJ/73PKzGxDm2hUZw+WxOuGAowO8G8/BdTHMGG5fXYcxcY7HlSQ35I6iSNF63QiEpJN8lOkSpuG5cSC0kHN+Yzar7jnpIZCjFUfEzpt43UrOgyBLn4sYTIv3HfgfJP3Bm1O6kkZFUXw5zbsSf1PShzd/+1pRFD/1unvkgiRpoWdwOOVxbR6TczizCowm/0uBEMq5urpaKTVeXV2NfQm5P6gA0oVJ8EIWeISimDcT6Xa72traipWFlBV//PHHevr0qY6Pj3V9fa2PPvpIq6urEZZ4L3z2OIBPgMCEzASWImiOrPjMBZsUG8/s8b2HCB4G8L8bFj537yOVhsXnwo0U3jcPF9woOCGMEXGuxcNH7snCK3ieer0e2QOUgWdzToS4mTHxdSm8A5WIrFLl2rwDz8RnPjaEPRQMeWiKEcnrF5B1emzclgFwvWDM6MrkRvTzLiX+RWX7DhRF8W/yc0rpL0s6svM/Koriy29w3crhUI1VWg5bvbCCfDmDR+oN+INAcb1erxcK1O/3A4ISK3p3GBqReMqPa/lCI9qKSWW9gkNdJrLb7YYCn56eamdnRx9++GEsLMEzffrpp1GO6gjHV9ghULwrkM/rEjyFB3rCoPr/Hjo4UriZ31vnyT07v0uqoDQ+528o/W1GBYV1b+y1Bk52cnA99mbAC3uK0aE030GJgPncz9dWeHEN6MtDDYyzZyc4j/dmPhyKwwthMOlRyDyzfBm54nsYbWQW1OFkuCMpN4AgJgzbouNz7TuQ5rP5o5L+pddd5w3uE9bYrh+eSSphPQrGhGJNiRGxspAtvnQTeA/Bg7IxKSg+HECzWfYZYFLr9Xql5zxC49ZZqlbtcS33iB999JGazab29/d1fn4eLcbYXoyWWhgt70IDgvHcdA4npXI1X14XAFlJYRbs+5se7uU9BGGeXBHz0ATv5eOF8fTu0Z7hAe7C2HM/cvF0QPZMCEaE+zhh6EiEdQN8jlK5wvHMyJyHYMgT3INUKq47HFaWHhwc6PLyMghnUAzGz8NMuh3lq2ox+L5lHIiH83B8Lg+Ljs/LCfwLknaLovgn9tn7KaWvSzqW9BeLovh7b3KhnNV1MojBdItOeg3v7JNAF5elpSW9/fbbKoqiksP32FdSCA9KyhJfhMhz75Tt4oXJ1WOZfTWfM7K8W7vdVrM5by6xsrKizz77TLPZTHt7e7E/wbvvvqvpdBqNOHxrKkcDhDluHIhDWZ/gMTcKlxOvuZddBOd5B+67CC349fw7bhTyOed7PKOnN/0aGK38u4y3G2MnA5vNZmUbLkhXZMG5FZ7VN0vx0AB549kw0Ow07ZwGkBxES1jpHh8U58ukHc05Gencg6eV89BAUuwEBTLGyNx2fF4j8Kcl/Q37/YWkd4uiOEwp/Yikv51S+qGiKI7zLybbfIQcK1CcIgqHifwOdGIwYPGd9b2+vo6tvpkAlBkrm6ennKyhOISSTthilB+CyPPVvqKQ+7lAI0xMVrvd1g/8wA/oS1/6kp4+faoPP/xQq6ur6vV6scwVJIEnkUqDBWwlO4BgMNnuORYRZzdzUFEWRzrfzvG68MHnkPMJ5RZ5U6lcSizNFZ68OD8TMi4vL1e2FIefgAxz5VhZWakU/zhZyngyVzwbdSLAf3aZRmaZaxApYeVsNot1JmRxSCe2Wi212+3oKYCi+joYR084LmQMhMT4YpyYZ99WnjH9rqQIU0oNSf+6pB+xyb2QdHHz8z9IKX0k6Uua71JUOYqi+HlJPy9J77zzTuFQ2eEQk4vVvfluZRGNLwLCk7fbba2urkZ7ZwaYNQOSgn1H2VEG8sO+Jtshq1ds4W3xSDwfi0wQAqzy5eVlVA62222NRiO122195Stf0dLSkt566y31ej11Op1YHIRHQcggKH3MgLaMAwLHQUjiZBFKgzHj+SGsciRgcxf/u0F5FTrw+JbnzON9jLmjP3rz09yTeaOZi3+H/L6Hel6M46EYz8IYMmb8DEHHZjDIH9di/wqQBc9CWIUMIFOQcygnUN1Tq8gI1wQNM2ceirhR5529pRpjD7/0eYnB245/WdK3iqL4zCb8vqRBURTXKaU/oPm+A//0TS6GN3DFJBxw7wU7jHBgwbGQ6+vrQcZB3vmgEyZgDByuUSeeUrlbL4uFPMUCOVSr1YKJduUi3Yj1xbDgYTxmbbVa2t7eVlEUkQkgfAERuFBiiDBa5LMdRuLR3Sv63gmgBngGBIp3dPSSG4OcFMy5gdsOJyu5Bs/i363VahUE5aEY3pv2ag7P2YWI6zsaY1wwes6040mlahiE8UCZ+U6j0QgkgjwiE/BSEISMv1SiWu86BCdA2TuhBLLhSIV7MTboASEIxoXwJK+heFWK8LVLidN834G/L+kPppQ+Syn95M2ffkzVUECS/qik304p/SNJ/6OknyqKYvC6e7iXYOK8qkyqLiZhUE5PTysbk66srKjdbscKPCw0EJesAvXf3pYbaMlOMEVRRNaABhF4Ag5nYYHkWHYECeV1JOPMtKfuPOZmgQrkIPDY43nPhGCU8LQoPQaPUAgegWdEOSCYvDZjkffw771K6fPvLPrd41mvU+A5kQfmkDJpkBehGGHh0tJSxfN5jh+PTOrVqy593JzsdeN6fX0dDWE9pUe6kIVp3huTvQcZe7gNKj+ZO0heZInnd84EvUAf+Du8lKdbLy4uKhyAG9tFx3e674CKoviJBZ/9qqRffd018wOhwGoyGVgvLDDLd4FwksLLLy8vq9PpqNVqRU+Am2eKmBougAmmv2CtNt8Z1nefBbKjbBBTXBflBTK6opKfZ6Imk0nE7xgSkAZ5bb6HMODZqGT0IheMm2ceGDcXYMIGT5tJqnhJ/uf6rKtHEd9U0d/08JoEf2aUHVSGAfTPZ7NZ9NPnnT1rk9K8ToOmsA6dQU4oPh6V5iGefvZwBX6Kg3CMJqNeRk4c/+jRo+gSxdgz1l5T4SGGdwgGhTBWPB8pPyeGgfxuSDEizq/c+QVEKDSei/iJgWNRhbOdHtvdu3cvWjgBw4CEPjh830kw7nV9XS7ecAXBenMA+9ybSqrEd84NOHT0YiMqyUhPXl5eRkvrra2tWLfArj8IK+9K8RPvANTl+os8qcNwlIr7597iVSmlfN7exFC40kvVPfWYHw//CAc9hOK7QGA8uufQp9NplH6jSMw9Y+IQmmfzwjI8Pw7GawHy9KPLBZubFsW8ozD9A7gGXY3w0hgNJ6s9O5aHJvliMZCIyzCIidoJnNx3hRj8og8Gg4IdfzmIQYQWBvb6el6njVV1PgFj4PXT/MtTbHhuJwidRIKgcyjWarUq+93jgXkXFBzyyGM2hMxha0rzjMHKykql9z2/D4fDgJUOmZlkYmSsvntHhN0F0hXOBe62GP9NFP5NzsmNAfPkfwfmEl41GvPNYrnH5eWl1tbWNBwOY+MXD9Oc1EX5gd8gN5AU8+5lyMyVFwZdXV0F249yXVxcVHpWOOkM4YphZi1Js9lUv9+vhL5kChgTCFBPM3sIhpyBCp3jot7FDSVyf9txJ4wAFhrrRo09mzjMZrNQepZ1OkmCUrG4wz2gF9cwqZPJJMp7W61WpI7Yk9DjVLY+8/gLpfbKMSYU6JZSitSSZy5QXC8CqtfnW4ixH0Kz2Qxkw3OwNyEQlRQZxkSqchQOBxEk4mWMjJce+1Jbnv+7Mc+LDof8uUcGalN1ibLy+dnZWUByHIWHGygOhgZFccOMd/cKQg8JGMPr6+tAha6gRVHo5OQkVhFijD3884yK9yMAQVLw5O9KOOwErjQPSVjl6obB/yeMAUm9aj7vRI9Bh275Z0wQggEvALyiZBfyzo0JawRQXLwze+tB6sDwU36MBQVucj12+kFoUWCu3Wq1KiEFhCNLUJ2gAoKDMprNZmxFRrtyQiPeCcX1uFBSQEx62PF8UrUjLfUOEH7AUDYy4Z0XHYv4gUXI4TsxHhhHN2J8LpUcBjIBWuj3+9HKHUPKfDLecC/Uk0iKz1zm4A2YN5wG/1hsRqGPjxfknhdo5aXKyAo8FMaXQh6MO8+NXHtoihPwzIWkyudkwTzj4WO56LgTRoBJlVTZghpPDxuMNQQt+Hc5D8vr6ROKkGjeSQjQ7XYlKZQLqC2VhBPWHLTCtT3FJpVKTykuAsEzeINRClAgr3g+JvH09DRIRQwB3/X21l7W2uv1QuBQej/wXO5tnUjkeJWwfFEHY+Uhl5NnGHXmdjweB8u+qG0440Q3aQwlpBnziyJyfRwHxofw00lpno8CIAhfOkqRRfLQolabb2mWk5AgWU8deyEP8sU9VlZWKpuvsAENRgSZJ2VOytEzRU5Y3nbciXBAKotIqKwiHeRbWEGyecEJwsOkcy0ng5gsXzvgNQnONPMZh0+Seyx2n2FwETRIPzyIN9FsNBpRaUanGmI2vARpSTgCvBXpJxAL9yUnDA8CYUjIwDhJCvjvJBseCQF/U0/uxOznPXKegAOveP/+/Yp3R4mlats0wiv+hhyAlAi1fA0C18J7MlYQgZLC+/u+FIzV1dV812vmkfkFVRJ+oqA8F+XDkmLXYke9zImPkb83yIV6GJwi886zIw+vOu6MEeAFSZOg3Fg0/xlLhxK2Wq3gDPAKwG48Iy29vYDGBQi2nliOCUAwMDzEfVI5MW6VscRAOz+HUmCpFDIMlFRWGrZa8zbkKL7XSlD5hhDioZzkwwOCbvLUkSswBKKHAa9T7s+LFrh+fp2ciPRwgM/IGmCsnQNyOSEswJnwnjgM6eWeiiA8jGRehwE/xOd4f5cJ5sHLuylCQ8mZJ4ySNN8Qln6FKZVl5ryPVFYMev0DCIj1Mtwfg49jG4/Ht87HnTACQNrT01NtbGzEhDIgQEEsOzAWJh9L7LAqz82S80cQXFi4F4wwZaKefvI1AfzNCSqalvI+PAdrDpgU4D0eg2InVg42Gg31+/2IPz2VRV4ay+5j5KGS171L1dZZPJ/DREdCnnm47fiiEUCemvR39nTx6elpIDAMuR/37t3TaDRSr9ertHmHewFh8T0vJsN4u4JJJZnK4R2EpNKQ1OvzbtKsLiXko04EA4CXdnnByFBv4mQkCAPjw5xTe4IuMJ7UlyDLvsDqtuNOGAGp2hbMWXGYTurgPb0G1EP48zp4FCJnfblOrVaLZZqSolAGJALMhlSC0fd0EhMtlXUAkD3Sy6QXkwYj3263de/eveAriEm9Bz6C4YaMyjfIRqncXtw9kWdRXJg4+LvzMt/rw4UYI+31DyAAGsHUarXY+AVEUKvVdHx8rJWVFR0dHQWq9DUDKCR9HUFYzDXhI5kcwi7SwSBQlJ1UJSsG2cOQbA5K7r0pHDESAvK+VDxKCkPlaWU+J4QljIDDcCMjVQu+vltrB77Qg5gUSMPg0zUIFp4UHspGagjiRSobbngqCEvvZJSkIJqA7FhfJsfhPMYCD5Ov8c5DC4yRVK7+414oHfyE72zk8XpRFNF5lnujLBhGvkMaMV/E4tCW77rhvGuHey7GDDQmKSpEHZpLZYs3CnYgmYHjOVriZ2TPDeUicpktzGn+ihPwQh/IO6lsauoVe5DATiTjyFgPg/FjDj3X72lDv7avsUCmFhG/i447IwEe0/HgTBSkDdAGOM5AeVpQUqRpJEVakHtgWRkgj08JPWBx8/pur+umnTTP6yXJXrvgAoD1dzLS4bukClRFqCCZiCNJZUklZMRQ8f7u/Tl439tSee6hvpeHP4vHv1574es8cBYYSEIkwjxfSehkH9/HkXBfR4qcv+jvvV4vOj9zH2m+FH4wGFQWe4H0kFmu53tOeJYLg839nP+SSuTK74QO/v68Q6fTiU1ukdFXGYI7YwRIfxGjra6uRvEOiu8CAksqKQyEwyZvHJrzCCi7M+lecuk7ELkQekzvKTgmxuNL7kmIwbbWpKoQZAwGoQYQ0Qk8UM9sNosUEwLt3gyvSAFTHi/6sYjc8/N/P488pcWiLsYEfgYl4X39XZkfmH9CAUIkruVoIs+1+6IrZ+uREb7nxTwQvr6K1TtXg77gs/KMllRFjSA37uPVnTgAN9zIATJTr9ejZ8Ftx50xAsAdqZxE924oi3doAcI5ySe9vBGl91fzFBPCQD4Xw7K6ulqB61hSX47rqTe8Oe+BwvK7VDaqdDKI64AgPEfMPTF4ZBekEiUgpAgQSuFEoAs6hgMFuwsKv+hw4pIx8Bie+fMuP04cezPQ4+PjyjJsxhNj4vMIQY0Corh4WBwCSg6xSJ2GVJJ8yCT39jUjZDiQE/62SG4wLFI574wN70RJMwVmyLsTxKDJRcedMQJeNoolcyjlHVxQXP+uCzmFFnhoYLynBJlQZ9CZOGAmyuK7/0jV5poob57jxeN7iAM7zXvkhJ3H5wghsT9xrmc8pDKM8j3vWAFJPtoNJM+fQ2R/r9+vw5/DCVgOb8RRr9ejDRyrMH0O8KKEg9Rp+Lgu4oicyMUoIBueMsRYOFqo1WpR54HsegiIQQG1IqtSua7DuQpQBcukuY9fE+fpY5ZSCs4Eg/OqcOBOVAwysKQ9JAXMY/24IwI8gMd7HCAAYDEDyiS6h5fKVJCTOZJiZyIG15EFYYYvyfXUlvcOYGNLFzY4C94Bj8TzuGfwUl7uibfxxUNS2eDC/+Ys8+vy+583//9FHSgYnpFxA3aDEBmLdrsdCupVet54E/YfGfNyc0qpHTV5Og6jDRpxOQCpgODYytwJQgwQxsNz/3nzE9/K3puqMpcYB54fw8c5ODqemed7VXbgTZqKvJNS+t9SSt9IKf3jlNJ/cPN5P6X0Gymlf3Lzf+/m85RS+m9TSr+XUvrtlNI/+wb3CIvmOXUUhRebTqfhoX1g+dlr8oFweaGMw0ZfWcagUSfuMRdeNyePuCcHE8m5UmlkEChgJRMqKVZAotA8H54jX9yDkeNaUtWLIhAOB2/jBjhexyB/Nw8MJAfv5UbbERBeFcSI8vNd5sR5Fr6/trZWicdpwMH7I4uSKujKn9U9NeXHkiocD9WuzLkbNp6dNRuSoi+gK7YrN6lq9IDnxTh4KzlHNh5a3na8CRKYSvrzRVH8oKQ/IumnU0o/KOlnJP1mURTfJ+k3b36XpH9F87Zi36d5I9G/+gb3CO/ohBhWDCV28oefGVRXTJTLPaELE/dwstGLhShOQgFhcb1k0/O+DjMRRid8/Bw8jIcRDiURWISAxS4gCAwWRotwAMHFE+bEoUPJRR7fw6nvJSLI7+XxuRswRwWEd8xdURRRF4By0ybe06EebvI9PndI7dkdJ1wZJ9Z9ULeAg8mvK1VrIDDsngGSyuwDz+FVolzbd8ty7oD7QBoTKrusEGbcdrzWCBRF8aIoin948/NY8x2G3pL0VUm/dHPaL0n6125+/qqkXy7mx9ckdVNKD193HwTfJ4w8K4LuLDBGgsHyFF0++C4shBK1Wq2ysow6bGd6a7VaCJcbJE/teKjhK8+A8S4MXv7LxLLhJOe4YeNd8f6LyE9PZ/mzuRFyQ7Bgfivj9ft95M/hnAkFUpIqRWEoq5Ok3umJ8QFlYtzd80qq7Fvhc4DSudxJ5SY5FHHR1wJn4ag0dxb53Pny7txg+DU8PPL3594eBrgBfVU9yLfFCaT5JiRfkfR/SXpQFMWLmz/tSHpw8/Nbkp7a1z67+ezWwyfPSTQ3Co4IpOpuOigKLCxW2eGUT4hfB9IvjzvdkwL73Et4qODP78ywp31uxi+eydEOz+cenutLqhSO+DvkFZD+bu5RufebKPqbnvdFHfl8+OGwNs94uPLzGXPv5d14dP7Od50E5vqegqRoTapube4kocsGiM+vhePw9/Def8yTpyjduYFwHdG4DDgSLYoiaidAgzgEN3SLjjfODqSU1jTvH/gfFkVx7JNWFEWRUvq2MGSyfQc6nU68PKSPTzgvxaorDjwqk0c9gVSFtwyYr67DwsI2Exc6DHMIiuBwXwQIEspLnX3AXWml6i5KfOb1DIyDezDuT2jg5KR7+Xx9vIcONu6VseWzu3C48ZKquXMPfyiTxUg6ySqVFYVcU1LMvYdOjI9nC5ATD+1y8he5ccLQHQKoAwWl7Bjn5srvz85zOaHI3POcPkYYOOaa3ovem9Ob2tx2vJERSCk1NTcAf70oir918/FuSulhURQvbuD+3s3nzyS9Y19/++azylHYvgOPHj0qmDCPp3wpMOk2lCcPH9yautfIeQT/uysZkI6JXF5ejqXLDhsdmTARDhX9OZg4h3geHjjyYaK4P8+bGywnx5hoBMW5EY7cA7xOIL6bx7drdDx8yomt3HAyhnzm4+DygMPIHQ3KDWKUSvlzRc3n1H/nWVFOeB7mCznOiTu+RzjCGGFgXL54Lp9rHEur1YrGtW4Qeb/bjjfJDiRJf03SN4ui+K/sT78m6cdvfv5xSX/HPv8zaX78EUlHFjYsPBBsBsdXd2GlEQI3Eg6HgeIMHAPL4KEknjkgLpdUaSuO0DBB3jgkh+Dc0603IQXkjB/cN2+KwvMysU6Uev0Bk+5xJdfIY0iuf5vSfa+h/6sOV+BFf8vDQA4fEx/kGUXNAAAQIElEQVQHqTTQDoc93ewID8X1uv78mXLSFtnxOcCQ+FoEQhFPLbqhQb7dsLtMcd0884QcS6ogQEJSFjXdxgdxvAkS+Ocl/duSfifN9xOQpL8g6eck/Uqa70PwieYbk0rSr0v6k5J+T9KZpD/7uhu4IjAA/O8NNDhyCOextP+MB3HEAFTz/C8/e7wOYsAAOJqAr+B6bjRQWOnl1k6ORBxZICTeGdZjRUIERwReyebez5XIhSX/7PfjeJN7O5fh6Mlj/zxkyD+nyCZHlR7qObLKw6MceudcjFTWXeT1DNwDcpYwEz6CxUeUE+ecDb/n2QqXFe7rRVHIFoVInEtPzs/FCRRF8X9Kum32/viC8wtJP/266y46PDby3GrOsvtgOTkiVfer92YUHHmKDWXycIO16p5Ldqjv8N4VEcX2kmCPVR0VuKHi8PfI430PM3g/vuNjwnMuOu4aB+BHroj54ePq7+GLcfy7KKbLDed4xsHDDC/tdS/rpFyeifD7SeUKPg9TF8kciMG30ONd3LG4g/T3ykMazoNEBinmHYoWju2bTdF39yAEcOVhQj0ccEGgYmqRFcfb5/yAVMZsRVGWbOLxGXRHBnk8dpsR4rogC2ef3Zo7CvAMxqIUlHsFDxlynoNz3DD58SooeJcOf848xufvudGTyhV2+dj4+DoR68bUHY9nG3wMHUn6c7iSUQDk8ossuIHmPOchfM5djvxaGCC+x+8YLK8uJEwhJICcvO24M0YgF17P5/sEuPD7+QxKvpLPFVCqLlHlIOwgrcOk+EQ6YUfBiRset9ZMOgLCvZyrgPRzWOecRG4MPdzByPk1/XBlygX+LqIAPxYZLA/PPFXs8iFVw0q+l1cC5mk7DyddfshIkSlgnvieGwLvd8h5OB94Gv6HM6LXpXt9r+mQqtxAzk0gP84NgQy874W3MPv/2jufGDmuIox/tWvHXu96vYSgKAoROCiXnMCKohyiHIHkYrjlRA5IXECCAwejXHINEhyQEBKISAEhcgFELkj8ERInAgE5jkPkJEAksOIYhHY9O+P1bsaPQ/fv9TdvZ9Y2VugeTZe02pnunpnqflX1qr6qV28WdWYBEQOB5ZQme+CjhB6Ll3XYrowec/vDkxpEtrTEbl0pKfaY1HO3DgTinZSxtwtnmZHgeFnM47FfCRiWeIAbulkKPssr6KIxKHlyLMC9oWnXT7sfd/cdQPR0nxthadJoMuuyohUl9mfoBh35Bdj2Wb/0JCVlHMixozLFXHos8O6FTXwXWBc1Ar5moOw+7dQJI1C6fb4e3t0haXILLQfhXNE83i5nwqWlphGpA2qSJppROCAX0axGk5q+9QwWlVqlUSndUj7DjOIhg88w055LOfO5YSgVYB7DgZLHMjRwg1heV3qE7jUtLy/nFnKMia8QxeP0mJ/fYUx9XYLH9VQx+nlqVVgWzuxcLjVPaXKfQ+TDdxN2T4Bn4BiGhyTIEgaIfosub7OoM0bA42GPkYmhpcmZE+/AEV83AG45/bMMHr/BgDDouG1YUdwsxwgoBGFW9hnLBdbzx45plCGK3y/f4ffk5CHCNAMw7dmW1EUvoKRpz6I8N8sAuBFGycGcpGadvdQAyJ5l8PGh4AaFBLzz3hBS0/mXVnAeVjJmvqS3lAHee7rSwUEHh71GBZ6RbZrXkPIGF5gFFksdwQQkTQwWA1POgJ5fLevwXaHKmRgF9uYSJZLMtTw4lN1rsr3ppQOA7m0Qd2LpEZayeMVntLLBpTTbELjg3YoyHxQydIHKsGYWn9OMwXg8zjv7EAKwEYl3E/IYX9JEebAvW0cRUS7kg1kWJfcQlWwWszHXlz0t3Fv1hVAu5z7zl/KOLHH/nPc26r4wTWo81VnUCU9AaiwwSuZxL622eTgouFvcEjj0kEDSxINh8Bx/SCnlRgxeKUYrKZqcUujjKTo+74Pqiu3ATOnyY9z8/TQqvZybpa4q/Sy6EWZRxussAnOAD4NLYw+Pkd2TKwFen7VRKg+9OM4EQAzue0dw7c7OTl4W7tiEyyndouiNEBE5TSlN9nx0IFBqjApeTOn50uOQ9PZBC4g6YQS4UQqDvE7ArWRZAMJNe0iAMfEZH+s6zQvg4eOq0draDQGGxj2JspzXY9ASa5jm/h8005W4gJ9bJPL7LbM2eItez4ErzOyPUaAfYBkOeMjpsuafBRB0D4VzyBaTxXg8zv38qPj0md2XMR86dCjXCDgY6BOau/+eLSvBTL5vPK4arGxtbWXlRz5pPDONOmEEXHl57+mgEmV1RcE6+4rBcpZGGb2HnxsLV1Kv9cbKU0PgGILH+j7IDKinGzEQPsA9TafS0DFzI9R0Wybt5jv8cr2j+g76unz4mhBXei/AkSoZouEHZeZ8txflEHd7wQ/88J7w0FvlcZ5JRdpv+Ms1IfDtHicGiz0QpaawSDq4aUwnjIDUKDM5V6r2PAVXVtN5yACq70Cgu//SfsQX48JDG4/HuR2Y1OxkxEMnXJA04X0wQ/k+BJ7d8Ov4Xn9fZgYWbcY/iGibnVLKzUOvX7+uzc3N7Jl5V57t7e3sRfKMfW8Ktv/G3UeuUGTPuPA9UoMbefUg1aVSM9GQ9uNabx+GwSjjescUeM29gjF4gRueDr8nVbLLczp69GjeiJXfOqhWoDNGAFS1dKlJ56C81GSDHbBQAzeRwXHj4MqMcGBdWT2Iy4dFXl1d1d7enkajkdbX1/OgYv29As1RXIwPg4rF98KQktwD6qkhz8psbm7m4ygyckJBDC22OMd+jHTg9djeU33lpOJLcfl9zyhJTZUrkwheprv5gNEeengJOjtrYXQcaPZMlIe33rmqlCmMznA4zIbNU4+zqBNGwNM6WEEeDEruKwSxlr4M02dgDy18diUMQOHJHUtNtZf3JoiIDO7g/knN5h5eHYZARcQ+lFmaXDfAvXqdAMfL5zLt+CKQYz9MBPTxZ3xGo1F24be3t3MbcMbAZQnMSWp2n15aWsrK4aGDhw3esQo5c9CZz4AJsLEJMsP4uruPbLiyAmIiZ0wieJhkLzzVyPfyxwREsRsyilcxizphBKTJNeOgpr5MFxCoRPyJsyRNKBaDRuEErpun51yR6WMvKe8x56u/fAA9fcmxaTF/Gf+7Mh+E1vZU0dWrVzUcDjUYDLLXNhqNtLe3l2vlCcFQEjf+7rKjWLwej8d5xva1HlIzqSA/KC0rAgHZkEvPMpAaHgwGOXz0wjavFXCP1QvfyGi4Z7K0VDWhvXbtWvaQ2CvT6xCkSn7xYGlo0vl9BzxG9yWfDtRMi/UxEJ46c5DRXT9CB3rCu9Lzu4cPVzvD+oyCUOHu4UEQ43m3W/jwHn/wPIvcaylpET0AhHxnZycbAHfF8Y5w8ZkckAf2XPC+/15RR5s4PydNgtMOKHuo6KEAqTzvG+Bh5vr6uq5cuaLjx4/nkAUizCWD4IVGKTVt0JwnZEya3OzGQXAvXd7e3tbGxkb2Wr06saROQdUIve+15+Ad5z3f638OuAEUkkIqv4OHSudYvAyASYAVvo9SYRcefhOeve7AydHenqYTwB8FPF6vQai1tLSUU3DD4TC37EYRR6NRrhtg7UeZydnZ2ZkorFlbW9PKykre3s0zUMiQNNmPQGo8EDf4hHmk9MoUHvLkVYFgEoDi3KeHip5aJBTCI+bzHPdGONynlzZPo054AuXsjWXEHfKb89TgkSNH8mDQFdjxgTLdw2yOErPYgusAJlNKOnHiRO717nXnjj3g2nnnoWn35v972k8YVGJqcvtUAK6srOQ0IbH38vKyBoNBVgJ2dX7vvWqDkY2Njfw5FM5lbGmpWc/vYQCKDz++gQ2KTPrYxx/Mwpenk+NHIX3WLjEutifjflzmvQsV8ikpg6HoAF4NOMhoNMr7OB4UfnbCE/DiD+I1d+N4+FKjTMTzKOZwOJxI9XENrpukidbQYA4+ayAcPHQH98pUjqf08Ey8/Bhe3XD0NEleA0DcCp6Dsm1tbenSpUva3NzU7u6uBoOBNjc3sxIuLy9nnACDAQhIeIZhwMi4Bwdij6KQcpS0D4OCt93d3Ry2IJc0EkWOCDklZQPgS8dXV1ezF0NWwfEmAGo8grJozcuQPQ2ZUsrGD6yEe5xFnfEEuCGstedlPcXms7znjj2l4l6Dx29YZS8dReFXV1fzNQwURSi4dvDCtQyWr1PgfqBFRvinkc9kPG9ee2oO5H9tbS278D5e3rRTkobDYXbFUURCt+3t7YkNaOGD7A/Kz+c89PMdqwgJ3ENk4jl27FgOITEoDnYju1545tWPZQ0MoQV8lFWm/A4GBEMKyM1khRE5qGIwuhCvRsS/JA0l/bttXm6D7tJ88y/N/z3MO//S+3sPH0kpfag82AkjIEkR8XJK6aG2+fhfad75l+b/Huadf6mde+gEJtBTTz21R70R6KmnBacuGYHvts3AbdK88y/N/z3MO/9SC/fQGUygp556aoe65An01FNPLVDrRiAiPh0RFyLirYg40zY/N0sR8XZEvBoRZyPi5frYnRHxq4h4s/7/gbb5dIqI5yLickSct2NTeY6KvlWPy7mIONUe55nXafw/ExEX63E4GxFP2Lmv1fxfiIhPtcN1QxFxX0T8NiL+EhGvRcSX6+PtjgHFG238SVqW9FdJ90u6Q9Irkh5sk6db4P1tSXcVx74u6Uz9+oykZ9vms+DvMUmnJJ2/Ec+q9pP8haot6B6R9FJH+X9G0lenXPtgLU9HJJ2s5Wy5Zf7vkXSqfn1c0hs1n62OQduewMOS3kop/S2ltCvpBUmnW+bpdui0pOfr189L+kyLvOyjlNLvJP2nODyL59OSfpAq+r2kjai2oG+NZvA/i05LeiGldC2l9HdVG+Q+/L4xdxOUUnonpfTn+vVA0uuS7lXLY9C2EbhX0j/s/T/rY/NASdIvI+JPEfGF+tjdqdmG/ZKku9th7ZZoFs/zNDZfqt3l5ywE6zT/EfFRSZ+Q9JJaHoO2jcA806MppVOSHpf0xYh4zE+myp+bq9TLPPIs6TuSPibp45LekfSNdtm5MUXEmqSfSPpKSumKn2tjDNo2Ahcl3WfvP1wf6zyllC7W/y9L+pkqV/Nd3LX6/+X2OLxpmsXzXIxNSundlNI4pXRd0vfUuPyd5D8iDqsyAD9KKf20PtzqGLRtBP4o6YGIOBkRd0h6UtKLLfN0Q4qI1Yg4zmtJn5R0XhXvT9WXPSXp5+1weEs0i+cXJX2uRqgfkbRlLmtnqIiRP6tqHKSK/ycj4khEnJT0gKQ//L/5c4pqGeD3Jb2eUvqmnWp3DNpESw0BfUMVevt02/zcJM/3q0KeX5H0GnxL+qCk30h6U9KvJd3ZNq8F3z9W5TLvqYovPz+LZ1WI9LfrcXlV0kMd5f+HNX/naqW5x65/uub/gqTHO8D/o6pc/XOSztZ/T7Q9Bn3FYE89LTi1HQ701FNPLVNvBHrqacGpNwI99bTg1BuBnnpacOqNQE89LTj1RqCnnhaceiPQU08LTr0R6KmnBaf/AsoH252bO7n4AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7a1pxXdE9Zjl",
        "colab_type": "code",
        "outputId": "6705f3e8-5879-4ce4-8eda-59b5f2defa69",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        }
      },
      "source": [
        "plt.figure()\n",
        "plt.imshow(healthy_dataset[0])\n",
        "plt.grid(False)\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s07Grpe9FqJD",
        "colab_type": "text"
      },
      "source": [
        "### Balanceo del conjunto"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AfUyislo9iW8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "idx_healthy = np.random.choice(healthy_dataset.shape[0], covid_dataset.shape[0], replace=False)\n",
        "healthy_dataset = healthy_dataset[idx_healthy]\n",
        "filesnames_healthy = filesnames_healthy[idx_healthy]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WKTPfcok-_cF",
        "colab_type": "code",
        "outputId": "b05744c6-8a98-4487-a704-93a20b5d1576",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "print(covid_dataset.shape)\n",
        "print(healthy_dataset.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(198, 224, 224, 3)\n",
            "(198, 224, 224, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cG6MSydopTqm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 80% train/validation - 20% test\n",
        "covid_train, covid_test, covid_train_filesnames, covid_test_filesnames = train_test_split(covid_dataset, \n",
        "                                                                                          filesnames_covid, test_size=0.20, random_state=1)\n",
        "# 75% train - 25% validation\n",
        "covid_train, covid_val, covid_train_filesnames, covid_validation_filesnames = train_test_split(covid_train, \n",
        "                                                                                               covid_train_filesnames, test_size=0.25, random_state=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NjsZMkRE_GFz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 80% train/validation - 20% test\n",
        "healthy_train, healthy_test, healthy_train_filesnames, healthy_test_filesnames = train_test_split(healthy_dataset, \n",
        "                                                                                          filesnames_healthy, test_size=0.20, random_state=1)\n",
        "# 75% train - 25% validation\n",
        "healthy_train, healthy_val, healthy_train_filesnames, healthy_validation_filesnames = train_test_split(healthy_train, \n",
        "                                                                                               healthy_train_filesnames, test_size=0.25, random_state=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iOTFyA7NtXMS",
        "colab_type": "code",
        "outputId": "bd853712-da4e-4454-fae0-f45432e7dfbe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "print(covid_train.shape)\n",
        "print(covid_val.shape)\n",
        "print(covid_test.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(118, 224, 224, 3)\n",
            "(40, 224, 224, 3)\n",
            "(40, 224, 224, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3wQKQ6SP_3I4",
        "colab_type": "code",
        "outputId": "11ee0b96-b939-4488-c0ce-d9ad3bda50de",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "print(healthy_train.shape)\n",
        "print(healthy_val.shape)\n",
        "print(healthy_test.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(118, 224, 224, 3)\n",
            "(40, 224, 224, 3)\n",
            "(40, 224, 224, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H1jkxsnaFwQC",
        "colab_type": "text"
      },
      "source": [
        "### Preparando estructura de carpetas"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bdL3bKbRuOQb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_covid_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/train/covid\"\n",
        "validation_covid_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/validation/covid\"\n",
        "test_covid_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/test/covid\"\n",
        "\n",
        "write_dataset_images_to_folder(train_covid_path, covid_train, covid_train_filesnames)\n",
        "write_dataset_images_to_folder(validation_covid_path, covid_val, covid_validation_filesnames)\n",
        "write_dataset_images_to_folder(test_covid_path, covid_test, covid_test_filesnames)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HPAds7VfBqtP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_healthy_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/train/healthy\"\n",
        "validation_healthy_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/validation/healthy\"\n",
        "test_healthy_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/test/healthy\"\n",
        "\n",
        "write_dataset_images_to_folder(train_healthy_path, healthy_train, healthy_train_filesnames)\n",
        "write_dataset_images_to_folder(validation_healthy_path, healthy_val, healthy_validation_filesnames)\n",
        "write_dataset_images_to_folder(test_healthy_path, healthy_test, healthy_test_filesnames)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oeQZo6wD0F4w",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "covid_train_dataset, _ = load_images_from_folder(\"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/train/covid\")\n",
        "healthy_train_dataset, _ = load_images_from_folder(\"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/train/healthy\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9JPxqUPXz9QG",
        "colab_type": "code",
        "outputId": "e3d104da-d90b-4d98-c5f1-27da766d2f25",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        }
      },
      "source": [
        "plt.figure()\n",
        "plt.imshow(covid_train_dataset[0])\n",
        "plt.grid(False)\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Iny7jLwZE_Yk",
        "colab_type": "code",
        "outputId": "c31ebfed-4cb5-4a2e-bed5-51c9e20818d8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        }
      },
      "source": [
        "plt.figure()\n",
        "plt.imshow(healthy_train_dataset[0])\n",
        "plt.grid(False)\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQEAAAD8CAYAAAB3lxGOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOy9WYxk6ZXf97+xLxlbLpW1dXc1OQ1Ks3EGomgC9IMEQYZkGBT0IkgGJG/Q+EUPNvSggR4Mw3oRDMmCngRLsGAJ8A5LsCEI9ggDCIYB2bNQI3KGTXKa6qW6KiszMjP2jMjIiLh+yPqd+N/b2WSzqtqTEusDClkZGXHj3u87y//8z/nOl6Rpqtfj9Xg9fnJH4ff6Bl6P1+P1+L0dr43A6/F6/ISP10bg9Xg9fsLHayPwerweP+HjtRF4PV6Pn/Dx2gi8Hq/HT/j43IxAkiR/LEmS7yVJ8l6SJL/8eX3P6/F6vB4vN5LPo04gSZKipO9L+qOSPpb065L+TJqm33nlX/Z6vB6vx0uNzwsJfFXSe2ma/ss0TZeS/kdJf+Jz+q7X4/V4PV5ilD6n6z6Q9Nh+/1jSv/GpN1EqpXfu3NFms9FqtVKapvGvUChos9koSZLMZwqFgtI0VZIkAs3wc7Vaab1eK0mS+Mf7+Qwjj4R4j6RPXPumf5/2t5tGsVhUsViM7y+Xy6rX66rVaiqXy3GP/PTh95KfC38P1/d//t1pmmq9XmeumaapNptN/PP3+Nz7PXEtvy7rwnuZ90KhkPkurn91dZX5fblcxvfm18IHa5t/r9/ver3WZrNRuVyO6282GxWLRZVKJZVKpZgD/uXlKD+3+XlPkkSlUinz/ps+e5O8+LVYM58Hvw7vzc+hX9t1h3lgrmycpml6kL+/z8sI/MiRJMkvSfol6VoZvvGNb2h/f1/Pnj3TfD5XoVDQ5eWlKpVKRlhcqFkESbq4uFCxWFS5XNZ0OtV4PNZisVC5XFatVssohyRVKhWt1+tPKO/V1ZWurq5ULBYzf1+v11qtVvH3y8vLEFoE7+rqSsvlUsvlUovFIr8Aqtfrun//vlqtlsrlsi4uLlSr1fSLv/iL+umf/mk9ePBA7XZbV1dXGcGsVqtaLBZK0zTm4+rqSqVSSdVqVdJWsFDKRqOhWq2mer0eBoZnXy6XrIFms1l8drVaqVAoaD6fazQaabFYxNyt12vVajWNRiM1Gg2Vy2X1ej1Vq1WlaarZbCZJqtVqWq1WqlQqSpJE1WpVtVpNknR5eamLiwstFguNRiNNp1MNBgMNBgM9fvxYk8lEpVIpDMlqtYq5LZfLIegYENa8UqlosVjo4uJChUJB6/Va0+lUktRqtbRYLOKZqtWq3njjDbVaLTUajXj/ZrNRtVr9hGFcr9dar9eZub66ulKz2VSlUonXV6uVSqWSlsulNpuNarVayANGK03TWEc3jpVKRdK1wvb7fQ0GgzDIvI/1wkhvNhtdXFyEXI5GI52fn8d3XV5earPZaDweh3xI+vAmXfy8jMATSW/Y7w+fvxYjTdO/LelvS1K9Xk/fe+89vfHGG+p0OlosFpIUk4ww4Hk2m40Wi4Xq9bqKxaKka0OCB6jX61oul6FMCHmpVIr3uFFhsp/flzabjUqlUhgDPEqxWFShUAiDUigUQij5HPeBQUGQpGuhvri4ULVaVaVS0c7OjubzuX7t135Ns9lM8/lcb7/9tu7evavFYvEJj16tVkNoq9WqyuWylsulKpWKlstlvKfZbKpQKIRQS1sBQlnyxo/nKZVK2tnZCTSyWCziWpPJRPv7+1oulyqXy7GW1WpVFxcXWi6XqtVqStNUy+VSzWYzlHGz2Wg2m2m5XGo8Hms+n2s6ner4+FiPHz/WfD4Pw8H7uYdyuRzPmKapGo2GLi4uVK/XtV6vY11Wq1XIRLPZDEXinhaLhUqlki4vL7WzsxPIADlbr9eaz+ex/lzXkQQGaTweq91uh/w1Gg1dXl7G/wuFQjw787ter1WpVHR5eRnXQX42m43m87kmk0k8uzst1s/0J3QDfeE9fObq6ire+8PG52UEfl3SO0mSvK1r5f/Tkv7dT3tzkiSaz+caDAaSpGazGQqDcvpPh9NMTB4xtNttSQrBwaJiSFarVQi+T7QrOYuBdcUQSQoBK5VK2mw28d3cS6lU0nw+1+XlZQhqsVjMIA2edbPZ6PHjxzo/P9dwONQv/MIvqNvtZp6pUqnE9ev1etw/1r7ZbIaHHA6Hqlarms/nqtVqKpVKqlQqga6YAxRHUtxPkiSaTCa6urrSdDpVo9GQdO3Fi8WiarWaLi8v4/sXi0UYovV6HegN41mpVLRarTQYDJSmqebzuc7PzzWbzXR+fq6Tk5NQHoxboVDQ1dVVZs09PATSIzeSVCqVAqGAxHAeGBA3jO5l3djhlf07mRueCeWcz+dhRFhT7nk2m8V8XVxcZJBdvV7/xDogS/6M3Cfyx3q588rDfkcK+bDi08bnYgTSNF0lSfIXJP2fkoqS/m6apr/zQ96vxWKhfr+vL3zhC2ENEThpa+VWq1UIAfCVuK9cLgcMqtVqAdPysbTHsSwCnhQLzfVWq1UGlroQepy5Xq/DO5bL5RBS4k+gOsZtMBgEJPV4+Jvf/KaOjo709a9/XQ8fPlSpVIrr8vzT6TSMAl4PZUYY5vN5zEehUFC9XlelUonnwAOhIKAYnne5XKper2s6narZbEq6NqjT6TSEkGs7iuCeWC+M4Xq91vn5uZbLpSaTiY6OjjQYDLRer9VoNEJBuAZry7qhHPV6PZSX66O8wPFqtRpGBMObJIlarVassTsMSbFG7n3dUWDoWUeM3mq10uXlpWq1mjabjRqNRhgU1hv5uLq6CiPlSlwulzOv5+8DJMK9cC3nz/D6bgSQiR81PjdOIE3TfyzpH3+W9yJIH330kR4+fKh6vR4Ph5X1BZG2xAcLyoTh9VarlXZ2dsLzOryUFApi95sRaH8/E+ukHUaAyUc4uGcUDuOAseHzCA8C0O1243vG47F+5Vd+RT/3cz+nn/7pn9Zms1Gr1Ypn73Q6uri4kKTwQEmSaLlcBi8CvMSgjcdjlctl7ezsxL3CNeBhGo1GGCMPh1xRUNqrq6vwmo5UeP3y8jJg+2w20/HxsWazmU5PT9Xv9wPl8Ux8l4c2UhahEBYQbjj3wPosFgtdXV2pVquFsWN+QAHwGEmSBKrysM1JO+6Pe8SxMK/wRCA6kEm9Xtd8PleapqrVahG+skYYFcITnFqe9MOYeCiAgnvY6XqB0fis4/eMGPTBTQMX4QVQfoTNLSEQ25WtVqtFXIdld2LLFZkBnMd7cz+u0CAFh1buhbDw6/U6460c+rOY3DPXB3E4ykBov//97+v999/XV77yFb3zzjuqVqvh5er1uhaLhRqNRkA/YlHpOk7nPaAl4nKeFeHhp5NgwGnml/sFJWBEuA7zj6BXq1UNh0MNBgP1+32dnp7q9PRUw+EwM9d4fve+0hZ6u2HGOwLxXdB5DpBHt9vVcrkMWWBOmN92u616vR5rDzGI0Uem8mHIarUKjoF7H4/H6vV6YbDynnsymQSyqlarWi6XgRBxInwOQ8szeagqKWOkIaed28EIuJH4UeNWGAFJAdu++93v6qtf/WrEhsBXFBlryd/cGzjRg0K2Wi1dXFxEmOCK7UKMQrtiMpHAfxcGBOLy8jI+LykUzuNZkImkyFZ4KMJz8cxOAl5eXurXf/3X9d577+mLX/yiHj58GKy8o6Bms5lBJXx/q9UKJeU58MA8A0YJRpv7AelUKhVNJhPVarXw7mQB3GCitPP5XGdnZzo5OdHR0ZGOj4+DFfeUmqMj0AghVLlcDuPkWSBJ4YkxSMgIPyuVSobJx3hUKhXV63V1u91AQawRqIz18BDSjQFzhvMAeY5GI9VqNe3s7GTWk+vA2Xg874aY9anX6/EaMu7ZBebaOQ9QkbQltj0M/lHjVhgBZzNHo5FGo9EnJsgXideBSCi2L5hDO+CrGwxnoT1FI21z4A6xgIEsCAME4Ax7vV6P7ygUCqEwbtCAq3wO6AoqkBRx9mq10snJiWazmT744AO99dZbevDggRqNRoQezmUwD+7dEQg8CTE0qEK6TrOuVivN5/PgBi4vLzWdTmPep9NpJuvAupF56Pf7+vDDD3V2dqbZbKbRaBRz4ASbK4OTXawxxpSQLR+WENMzR0B1frp3x2i3Wi3duXMnw9DDPYC+mCvPFuXJaZTR+YbpdKqdnZ0IVXgfxofMCcaQEBHjBmKr1+u6uLgI2c4jhTxP4A4RecMQ/CuFBJhUBKLf7+vtt9/WxcVFBnLjEdyTSNs4yK8lbZVIUqTjJGXSPVhgZ2/5PPF8kiTh/SRF+s7jYNhzoO5isYiFhTRysgtDwGJvNpswHlwTA1UsFkMZue7R0ZHu3bunO3fuaGdnR81m8xO8A8+BknA/6/U6hLFYLGo2m6nVamWgPUrm6UAEFyOANyar8dFHH+nDDz+MvPzOzk5cF+TlaM69qxOurLHDatYDwQfxSdu6B7xho9FQmqaRqSBz8MYbb6hUKoWSgSJIB/JdeHspyw+wDswr68YgC+Ch4sXFRRhpR2E4Bb6DOWW+Z7NZhjPJZwKcAPQQBsOGLnyWcSuMgKQM9Pv444/11ltvZeJtFyBJmTy1pFBACCeHp5IipwxRxPBFJozg71hwRxcUprDYoBDuCS6CIpydnR3V63WVSiXNZrMMhM7XQcBhOAHJwKBtNhtNJpOAg8+ePVO9Xte9e/d0cHCgRqMRcBpvJSm8JILPnDNHFOrkDSQGqVwuaz6faz6fB3IYDAY6Pz9Xv9/X06dPdXR0FMrebDbV6XTU6XQiAyIpQjOQnaQI4xB6npPX+T/3Sp4e8s3XEGXC+EpSt9vV/fv3VavVdHZ2pul0mlnn2WwWyo8xcmOa97qSQo48hPA0I4oNonISlqwK12P94Y+QDT6L8WBtPu3fTZWUn2XcGiOAJwZijcdj1Wo1LZfLgIQsAJPBYIK4jr+X15wcy7/HyRcMEQro9QAofqvVClTB5xAIZ2tRAIqD8AoQfMSTjkCce/AQg+9wUm42mwVamEwmevr0qZrNpnZ2dnRwcKBOpxPZB+C4ZzW8mATj5NV4eCaH2igxinR6ehrFR8ViUd1uV61WS51OR+12W/fv31epVNJ0Og2Sjp8I+mKxCOF3rgcOwQlM6Vpp/Tr57BCysLu7q263q7t372o+n+ujjz4KRcH7gwhwHqyVx+MooUNulztk0ZVZUsbg5lPJ8B8eNvqzMJ9OhjoaYN0Y3Kvrwmcdt8IIOOFBwcjp6akePXoUnpKJciVl4hxaMmC1/f2w9l6owXtc0bCmFJg4a8t7HBFwD+5h1+u1Wq1W5Njz9+RZAEIWwguP7Ri8H7SBQUBR5/N5GJlGo6Hz83PV63U1m03t7++r2+2q0WhExgBEtVqtNJlMQokuLy81mUw0n8/DQ8LT4O3Ozs6i4tFJVuB/o9FQo9HQwcFBhE3wFxgjL/UldicdjBFwhl5ShgfCgLA2PBPzs7u7q8PDQz148EBPnjzR6elpRl4wpswBMB9ngVw6MnGCLn9PrD+oCdIxbyB4Ni/9Rr6dkOX3PBq6KROQ//fjjlthBCQFPOPBhsNhQC6Ug785G+0pOBbNYa6z75VKJYoyHPY6YQcSkbIIIZ+npWYcQWJRna8gr71arcJre80AsB3hBUrmF9c9j38nSoMAU5sPa06dPvFxq9VSvV7XeDzWdDoNYZxOp8EZjEYjDYdDXVxcaD6fh3IBYZlDNj8RqrXbbTWbzSj8IV0JMsKjUTeQn1M3uKyNrxGywPrzfowgz3dwcKBut6tHjx6pWCyq3+9HcRXZFAwXz8yzsbbuPOAoMNweGjqq4hrwEM5J4CxI5znvwD++zw2ghyHIIbwI85OfK3//Zx23xggAax3yT6fTyP07PPfPSNsST4QE5XK474ypp/N4nQXGaOCN8baeFkKgvWAEA4JQeYyJsXIUgIHA6yA47oX4yfdwz4Q1eFBnjPGoKB+1+ldXV1Gfv9lsAp6TNSFOPzs703g81nK5jE1BhUIhiMdarRbKvru7q3q9rp2dnUiHeTzPdXmN/QIOkZlvoLN7VofEzk9IijVkXuv1ut544w298847arVamkwmOjk50WKxCEPssrNarWKTmSs4ygzKgVdwMs4zOhC33IekIJGRA1d+HBLvRRYwyKSykUEyGS6voAWMhKcwP2uVoI9bYwRQRmdqz8/P9dZbb0XZKcPjP4f7QLw8H+DwWdpuNvK0lLSNqxheEehpLWeype3GoKurq1AcOAR/NpQUWI4Ce1mxzwEK4SgHw8HiSwpyiTBHUpB3CKm0DVPwUp5rLpVKQUpSacm+B+aoVCqp2WzG9udGoxEGAPjrUJV1w+uBODwTg4FEgN14M9+sY17YqZpstVp69OiRfuZnfka1Wk0nJydhIImznblfrVYaDoeBJjzM8IyEp2td+Zxw5W8YK57PKwO9pDtvjDAgpGWpMnQZRjb9/3kU4Abhs6YGGbfKCPgDQA7iiT2Gl5SBZp4qdG/ihB2vOwOL0pCP9nDEEYK0JRFdUB2e+sKQp3YY6egD4UTRHXl4GMB98jsEH4U03D8IihgZwqxYLGYIL17DGHgIhHGBxCOM4PooE0ggTdNABxhXoC/PiQJ4ypE5lLZGScqWXSPcvo6sNfsDUKhSqaQvfelL+vKXvxwhUJJclwPDAX3ve9+L3YHsdkRZfR25FzZIuWfNGyM+Q+jlBsZrHZgXDwN8OzGkKlvP/fndCLoR8nnldQ8jftxxa4yAKwALj2elFBVBdFjv1s+v4bwBVtqJIHbXkcrzqkMvXuEzGCKUieGkjYcTpAV5DYIQ6EtFW96z5KvRPKYG9UjbMAGhc8+EsksKxQDtcB2e12G4hxzMIQIN3Md4enUj13WSlHQr184LKGsDN4IhQvjde2JAPFMD1G632/r5n/95FQoFDYfDQG6tVku7u7t6+vRpZDJIrTr/xBqyvqzPfD7PrLVna5ARDIin5fiMIxrWk98xBCAAT/1x/TwnlFd+H84VvAg5eKuMgJRViqur622xpNmcF/CUIpMmbb0MguXxvNer451pzIC1dsG+CRl4xaGnNInpWChvcuIxNfG5tIXOLoAoCx7WEQMhD0oCfOd+vUpO2nov7oe5hdhzdp3vcIjstfTOe7gxdIPpxKaz+ygyxtrnzreDEx4Rm7MGoD6MAfNdLBZ1eHiovb29qKhM0zTqEw4PD/XRRx/F86NwPLOn3/LhGGEKHAxzx/y4rIIkvSkJ888a8zlfU0dJDv+dpLwpDHaD6ArPaz/uuDVGgAeEdeV3FlbakmVO4rGQDA8NECCUn9ew4MBmz5FzL8SEDmMlxfs9PODzno2ggYXDbMhAJ6BQAkdA3ANGjGd3oXCv4GWqCDGKnv88IQrP4p6eVJVnVhCq/ByjJJBX0nYTDyw5Xt7n1WNYT8e5crrxgc333Y6eNSiXy9GN6ezsLNJu7MqkDJfyZU/3eTjmKIhnpXqPeXDEiXPheSERvcJys9lkwkyXjTzC9DlmnlzRPVvgis5nbjIKn3XcCiPAxDJJKDjQkuYYToThDVzQpSy0Q8jwnCgAltmbljj84neHfSiV1+cjpN4JyVEKxgdDVCgUYuuqC49zCk5YOjT052POMD5OQt2UyeDzTiZxXZqGeGMMjLAbAZ+TarWaidVZG0cRhCsYMycZuRb3hHB7poPYn/ehjP68zBNVmWdnZ+r3+0HODodDJcm28YjLlMf2niXyefbaAS9nZu0cqiMXLmOgGeeNHNWALF3J8zyAtCUFHZH6/19U+RkvbASSJHlD0t+XdCgplfS30zT9m0mS/OeS/ryk/vO3/uX0urfAZ7lmxlswwdPpVK1WK0PSuIDiwRwpsHCuFHwWwozv8HgvXyPAtRxxAAFhnckD+x52PBqfw4uRHaDVFvULCGWepUaZ3Fv4phJpqyAObTFezIP0yaahhBQYLDwYyuw8BSGAG7d8iTSvo+yEBBB5wGFHTo4UvATX421egzcgtLhz547efPPNuA8Itk6nozRNY47b7XYYAjfqzEWj0Qh4vlgsMnKVR2nOSTHnGM+85/YiMg+B8k7CPwNi4D155+QhHqjoRWoDfLwMElhJ+otpmn4zSZKWpN9MkuSfPP/b30jT9K/9OBdzEk/awrL1eq3xeBzVZ04oSdmMANdxa+oog4kil48i8F3E9x6nE1K4QpGLZ6MMC8p9EPdRHcd3e5kwhsBLbl1QXJHcyzM/HhI5ZM1nKrj/PNm02Wz7IbrRdaPHe/DSLqyEANwPZby+UcnLtPk/iAFUhfHxTI9zInmUVK1WI+bf399Xp9PR7/7u7+oLX/iCGo1GVCyyjp4F8XlGcX03n8ucKyMKjbwhZ+xVcfLTvTeGjrDJMxr+Xkd+0pbX8hifz+YRgPNIL2oIXtgIpGl6JOno+f8nSZK8q+tW4y80HApjQRm+Iw9F9XpuF3CU1VlbVwI2wjh0JevAe93au/CweN1uN5BEp9NRtVrV+fl5fIeTZNyTtN05CO+B52IzDe+dz+ex/9xZf2JlOi8hnPzOfGCEnHD02Nu9nfcH9CyDG1cISKoBUVg2dcFxkM0hRKhUKrGPYDqdajQaRdbEiTnfqOUcjCtSsXi91353dzf2IzB/v/u7v6tms6m33347KhjL5XIUPoG2PN3MmlIMhFf1sJS18Ln0jI9zBdK2gAlOxDe0OfPvmSE2lZGGdePNHPBeTwvmQ4MXIQQZr4QTSJLkkaRflPT/Svq6pL+QJMmfk/QbukYLgx/xeUnbeN5j2TS9TjfN5/MQQkcBXrCDEPsGkLzHookm3hsj4FkFDysYWHpP56FAvV5PjUYj2qU7IYiAEPs7FOY1Uld0CaLQh3lA0d1LM28om289pTkI8asTjHgNYuXNZtuP0Xvw8cyFQkHtdjtCAfYQUAZdKpU0HA7D+FxdXandbmt/f1937tyJMm3PlHj44whNyjYB4T6azaaazaZarZb29/dD8GH9y+Wy3n33XX3961/Xo0ePopvxZDKJtKE3MJW226JxDO6ZUXqKm9gP4Khns9m2EsMYQbB6KAVPxD1TAUiDVuQD5fZQxP920z+u65zYi4yXNgJJkuxI+l8l/Sdpmo6TJPlbkv6KrnmCvyLpr0v6D2/4XObcAZQPa4tQMWmz2Sw8q2/BxfO6oOPV8JxeIESHWIdirhhO3nBdPIOkYJvxMEmSRNy5t7enZ8+e6ezsLFOVKG1JKQ93ns9DdAaGUfcUnceP3AehSLlcjsIdaSs4eFuIKQwkDDasNfvWx+OxSqVSdC2ezWZhVNI01ePHj+P/GGP2DrAG1HLs7u5qd3dX9+7di0alIDdXDNYCRXB0xBo2Gg31ej21Wq1PlEczF8xdoVDQN7/5Tf38z/+8isVinD1BihD04WvNM+W9q8sh4R1t7zxMJJTD0Lm84IA8tcj3eFswkBDGyTkdrw1wj8/veYLwRcdLGYEkScq6NgD/XZqm/+D5Ah3b3/+OpH9002fT3LkDKBkK6EqNR80/tFfK+aL6gRzEtQg+Mbx7dCbQWW2MgAsIcfHzZ4v7oyZgZ2dHDx480N7engaDQdTJO18hbevQHU4CrT0upagIQ+XNKYCVwET65fEcm80mkBNsdqvVik1EpM0Gg4FGo1HsMyCf7odm+D55QiU2MpEK5fm73a4ODw+jaSnlsMlzFp/7hSD08I3UWaFQ0N27d9XtdgMBQYQx7x66cd0nT55oOp3qrbfe0mg0ivMMUFx3Nr4uGCHex3e4onk+37cCs5b8pGbAS7txLIvFIvoNIg+u8O5wPNR1eXPndRNh+CLjZbIDiaT/RtK7aZr+V/b6ved8gST9SUm//VmvmSc2WKx8vArczbOnHla4VwFVAN8wADR/dMLIY3qMhdcRkK0AIvJeYuJms6ler6f9/X1NJpNos4XwI/gulBz2Ac9wcXERNfl4e56DRhoI4s7OTsSSzgt46tEREnzB7u5uQOzhcBjxM6lXatgdTUlbdp3vhSilh4FvuKGnn3spWo4xHx6aNRoNNZtNlcvlIPeSJInyXxTBQwa6/BK2PH36VP1+X+PxOBAOyunFSTgPzxpBUpIeLJevuzPPZrPgjDCu4/E4wiGUHJnyXg3eg5LKRZdzDLzLnhsPhpO9GIe8kXjR8TJI4OuS/qykbydJ8lvPX/vLkv5MkiS/oOtw4ANJ//FnuRiKgcdz4gtoDxTyGnhgs6eTpG0qDSVAeZ1vgKQjfnZk4AVLXJe++97IkjiehaR2vd1u6+DgQPfv39d4PFa/39fZ2VnAeITFF1LaVqF5HO0wmM95bwTmQspW3fGs8CIoNnPY6XTCyEGSnZ+fB7x2xcvHucDfw8PDDGfheX/2GrA25+fnOj8/j8xMnnSEoCXswAhymIo/Lx6XeL9ararfv85Kf/vb3w6Oh/vlmR3BMbg/5MqzFoSkGH1alzk8zxsUOgpL2xCNZ8bLE6LexOyzRs6ZOAJ2/sANgyOFH2e8THbg/5Z0k/n5TDUB+UHZJRPjhB3WmXjXLbrH3kwasRjQar1eByzlWtRtY8mfP1NcC8VxA4D3qtVqmRSYpIwAEff1+311Oh3dvXtX9+/f13A41GQy0Wq1iph1MpmEgSNtRpUcuw3xMggTUJPnw5jgwfD2CDPCgtEsFK7r7D2u5rnefPPNQFKk/OBlUDi6Bt25cyeMKUeRcfgG8ylt27YNBoP4Ht/Zx2dBMfwdwypttx0fHByo1+tpMpkEmkrTNPoo9Pv9MMoYDuk6pPRKPWm7LTuxFKaTzrwO2VkqldRqtXR0dA10UT6QWr7ACQPi/IDv1YAHccPgnJCjAXcEbhA8XHkRAyDdoopBj9E9N8yEcACkk4C+n0BSJmZ3i4nASYqfzWYz4/E8nPD4cLPZlsd6BRjhhvMReAuM2WaziUYd9+/fj3Sidxo+OTnRP/tn/0ynp6eZdCRFRcwJyMY7J4NY+AzxL8/Ne/IbYSBWZ7NZpqlGq9XScrnUvXv3oo4Bbz0ej+OeuIZXSLLDEUOF8cWgYoh5newCn0UZ4TlYB0Igvuv09FSTySTCLtaURijlcnzwCUYAACAASURBVFmdTkdnZ2dhWF3B/KBYDCdyw+8YUAwf3EqxWFSn04nnAK3R0Qm54dreUAQHglwgUzgO56CQM4wZSMNDAecuQGz/ShsBJzhQfGlbV+0lsPx0yOueErIQQYTw4u8IsTd7JNVG7OZEDwuH5yyXy7HVlpQaHoyUFcIMNL+8vNR4PNajR490//79gJRJkuhLX/qSft/v+3361re+pffff19Pnz6Ns+q4L3L8GA+v1R+PxzEvEJTSlhfBsJRKJR0cHGRiV8IjJ2Vh+TnxiPvc3d3N7AxkffCgXIv6Bog/eg04Qeh75lljvgu04evCPaPAoDKUl3ZoKHqtVlO329VgMPiEo2B+PKNCGMJcgfyc/EPmKEnudDqqVCrq9/uh+MgB38O8SluDQ5jAs2OgeC+6gPI7Getcg4ddTiS+yLgVRkDKHq0kbReJOI6YinjLWWUWyT0RXpB996PRKLyXK4cztEAwT+M5a813QYjxXjwyf0PQgckYpk6no52dHQ2HQx0fHytJEr3zzjv6whe+EC26f/ZnfzaeBwGntgEPsVgsIl7GS2CsiIW9tJhc/sXFhfb29iLWhXB1QhVDSvdhz6bg1aTtgZugDG/Rxb0T27uQupeTFOGOG3NJ6vV6QfqRJfGt08PhUN/73veClGSOCYUw9IvFQu12OxM/u3NhvbyC0TkDGHv+f3Z2Fq3M+N7BYBAhUD6Od1YfBMd3IJ94d0d4+RQg18l7fM9ivOi4NUYAr02Mh0CwMCiUCxjWFAX0VmDOE8DuI2BMNlCVFBseAUZ6Op1GjQIQH8Un7keAnPVH0GitxeD/H3/8sZ4+fap2u63JZKLpdKp+vx9GDmNSq9XU6/Vi8RE0DvI8Pz+PPPlkMglvAiMP78F9DQYDXV1dqdfrBfkKAYfRQCELhUJ4b3bH7e3t6fz8XJIyffpYp3yYhlFG+TkOHGRWLBYzfQlBONVqVZeXl4G+4IbI5kCu8vrFxYW63W4YI4jGTqcTnju/dqAez6BAcLqhgBNhnfv9vgqFQhzCUiqVdPfu3Ui15uN1L0t3BfewQbr56HFXeP/pHBaffZlxa4wAAoQgEyehsA4BgeVAfYfeLAKlsU6mSVu4hXCR7kI4ms1mdMGlZRgCi6BI277zDre9oId7QWBBMKTjIMAGg4E++OCDQA5UDmJgQDyk0yjN9UKqzWajdruti4sLTafTIJ/w1I6YKLoitQdqci+OEXFeYbW6PuOAMwSazWbMIcSiw3bmmrUsFAqhJKxfs9kM484pTWzi8UwGxqJarcbx9R52FYtF7e/vx/fBARGTewzupK/Xoji7DuG4s7MTBUFUZuJsWFPgOmiDvQTOXXnaGQPphsHTx44CeI+Hp3lCkHn+vUoRvvJxk7XzxcnHPh7XeayHIOVTLXgCqufwZqSp8Ii+MQSCR1LGCPBZ0nZ4VP+HYi2XS52cnOh3fud39OUvf1lf+cpX9OGHH8axVeShqUjDINBfn1iSYhNiyNFoFB1qnED13XAeixcKhegQTHjhO/8g4JhHT8Ni7HzHHPc/mUwCTmOsvLaBcIo5I4tAkY8rCkab/2O8aZJKLOyhw2w203Q61cHBgR49ehQcAV772bNn8SxwDPlncYiNAZIUBqhSqWQU3Ju5SteGrt1uB/nIM/Ac3DfoxUlojACOzo2ThwPMp/90LuFFx60wAkwmrCyKjmAvFotMbzs8LBANRZK2Nf4oPdVursAohKTIucNeA8fxtJ6qIZbzdB0KQmxH+e90OlW5XA6CrVKp6MmTJyqVSvra176mr33ta7q8vNT777+vBw8ehPJDpnkueLFY6PT0NBAAPQC8EAZlwnAwDxBR9P3n2sS6hEXtdjtTkQgq2d/fD56B8wkwuuPxOK7JmjlPwVzBMTCn7XY7sxmJZ0YhUTQvJqpWq5mmIZCnoJ5+vx/FT3hp6bqnAMaT+eQZnYx2pXTDiNK2Wq0gljG8GC9Cx3a7HYQs9Rx4d4ZvkPJt1W4oUOy8IfDaDdcdd5ovMm6FEWB4DhchIobkUEwEwFltPuuxLaGEw3ksqpOCQGqYZlpvo/ikyjxUwVKDKNbrddTk8ze8NQu+s7Ojdrut4+Njfetb39JXvvKV8Jx8r+fz87XjeNfd3V1JilbakjSdTjUYDEJQpO1eBe4bY4g3gkREqSSFp8TTjUajTAHQwcFBpPbG47G63W7MNcYAxOIG/fj4ONaAIiVJn+iuxHOhQHhb7m0ymUQ4AoqpVqux3+Hy8jIQCAjv4OAgDMh4PM40EfWYm/tzI+H1ILwXo0rpdZJsOwvTD7PVammxWGg8HktSxqg4YYvyEgpIn+QG3Ji7wnt60AujXmTcCiPAwwHf8ykX0ICkgJDSNSQiTSdlm0BIimIgro3guLHJp4rgAXyjkafVMAyctjOdTiMO9/vwgyspCFqtVtH26jd/8zf1zW9+U5VKRe+8807Eoev1ddlwu90O4SiVrttlgUyo9V8sFtrZ2dFisVCv19NgMAiFaDQaERZ4zT3zikdiE5CjoEajEduZvQvu6elpxjMyH9VqVdPpNNMKjnTl2dlZGFgUioNmmVfnPbg/j6mpUCTFybXa7XbcH4pLNoPwAc86Ho+1v7+v8Xis0WgU6yspypsJC9P0uu06adHJZBKGC+X32gA4gvV6HacTw+uQDZG2dQ8gDK9RwWjzN6+VYU78844O/rUhBknpeNzHYrOQtIuGUHI2GiFw9hmFQzgRODwQFpXSX1DDaDQKL8a9Ua9fLBYjNQTp1263JWWbQACtWTxaXWFMPv74Y52cnAQhdu/evSDICHsQJK8r4CxAYnLKrNfrtXZ3d0MogN7Esj4/FMAwJ06YrVbXx5JxeCfPADnmaMNDEVBEv9+PrcebzXWhD38DviP8rJVvLW632xoMBsGzDAYDHR8fa7VaxZ6MYrGo3d3dWL/d3d24T9YS0pNKRPZcUPmHImNAuCfmhrBiNBpFVqLVaqnX60VoSDqyWq0GL0Jal8wO+ySc7PPUIMYY55XnChwVoBfIFJ9zRPwi49YYAeJUt3xMnJMlsM8w4kDv/GDyYKHzaUO8BIYBqEeqDQadVBTkohOQbsHr9XoYG4e3fM9qtYq6efgHDFCSJDo+Plaj0dCDBw+iHTq8gkNUoCweyTsksReAUAV0Ajx1Vhmm3jdBDYfDeN7JZKJerxfCOJvNMsSl5+2l6/gWha5Wq9rb24tn8O3LhF4YJIwgxny5XKrf7+vk5ETL5fXpSfQrOD8/19nZmVqtlo6PjwNdvfnmm2q327p79646nU6QnoQPq9UqmpESZsAbkKkBZRWLxThYZTweazAYhLHzA2MgLTEc1DQ4j9But1Wv1zUcDkOeMCAYcq8F4B/D4b//zUMBJxFfdNwKI+CpIE+L+GYij6uA6aADabuzkFjcvbDX9RPfM2DK8WZeeovH83tAIb1eH0+DF5C2/AZKCzno24vZBLSzsxMpMO753XffDdSDVysWi2q1WjFnkjQajYIo9PiWTIGnXPF8pFohZInj0zRVq9UK43d+fh5wfH9/X6PRSAcHB5pMJtG9l7QqJJf3E4BAQ/CB6My5F7pQxsxGKz/9uFQqxf1R68BRY5VKRe+99556vV6UDGNQ4CwIBQlvvPCM8nF4kyRJYlcnZcNwHCA/6iUwGJQrY1iZ/6urK+3v78f5i5QiI1de4OV8BIjgh2UIPG3Ie180Q3BrjIDDIEmZuIhcrJeTkhUgZkbZ8G4MICeFQAg4qIIQQFLwDpwRgDeHZOLzHg8yMBze48DRSLFYzBQXVSrXp/reuXMnFJvYtN/vR0s1FhxWejAYZGocKMjBI/hedZ4Nz+gsPlC91+tpPp/r/Pw8GowAnT1v/+zZM9XrdX3wwQehtBgAJ+ngSjabjQaDQfAyXqZNCpV6DOZ+OBxm4LNX1lEN6M04IAnhYJIkCVg+HA6j4WiSJOp0OrG+oMk0TWM/BwgGYpTuRE5YIgvMDd4ctNZsNjNyxXq4keR6bGtHiT3Od9ISZXeSmc99GoL4ccetMALSltSDkWUSJWW8L0U5TBQbS1AEr/13yOkGApKInWUOu7w5qG+QYWGpI0ARUTQKQaTskVpO+HE/xPsPHz4MCI/QoxSlUkm9Xi/SZygTEJwGKRBTxMJ51hsFgosgdKEYptvtqlqt6ujoSE+fPtVoNAq0QpiFIfVMAiEBBq/T6ejevXvq9XoZPoeDY5rNZnjx5XIZJb0I92g00ng8zsTP0+k07hskRvhxcXERToGDRvDmXlRVKpUibPRaBt9pCpICQaxWq6jBYM2kbUaGtfRslGdcKAFnXwf7P5AFngeZxXDzPXnYzz9Pn+K4XiY1yLg1RgClIcfvkMor3rx+3KvhUAA8JEIrKaMcKNXFxYVarZZqtVqm8ShNMbgWcLZarUa9PPANQsszEtwf9++QDcPR6XSiAIXP8b2FQkEnJyeq1+uBSGDaaW9Oug+UhEDiaZ0/4P/wEDwbP2ezmWq1mg4PD4O9n8/nms/nUcTkho/nobCoXq+r3W7Hjj4MBaGFb5iBHOU1Xuc7PAS8KQ52UlK6To2yrRkWf7VaRTUj9QGDwUCHh4cZRQQNIFOkikGUns9vNpuZAi2viQCtIFvr9ToyNoQ9yCFIBsPnBgEZcQIxH+u7s7oJHbzouDVGAOG5KSxgwsmf430oBwYBSNl2zaAHJ/BQdpAH3sEVFY+O8hM+EFOmaRrlpEmyLbP1cw0RVgwUyAYEAnTEU7h3csJzPp+H8g+HQ02nU52cnGi9Xuvu3buZjAXenfthLiHiMLR0HaLTbavVivh2b28v6jLSNNV0OlWv14u593vd399Xr9fLpPm8h4G3EKNrEUaAudtsNhGnM1c0XvW9914mTWqxXq9Huna5XMZPzh4gxj8/P9d4PFaj0YgyZQ+LKGCC9KQWwslgCsr4fnZbIi9e5UgnKVAJn+F5vdBIylb7oeRS9sQmT3/z71UZglfRaPQDSRNJa0mrNE2/kiTJrqT/SdIjXXcX+lPpD+k47A+OEqOYPLjXD2BZWSBP53mxhfMKkEN8BySdx2DEg060kGYjfcZCs5AYH67j5CavAcHxWHw/3p9nIOSYTCaR3hqNRtpsNtHYlHu+uLjQ06dPo2Z+PB6rULjuCUCbL/gS3xxE+NDpdEJp8wiMXoR4V+J3FIeQCbTiFXrEydK2whCSjbWDa/D6ABCGl85St4AhADVQx7+7u6tutxvvJ8So1WrBDfDsNE8tlUo6PDyMugDmQNoeQkIfBYwNIV+++o/QwTM0l5eXseHK5xROhGdCfjA0XiyUJwXzw5HxbUICfzhN01P7/Zcl/Wqapn81SZJffv77X/phF3AiC9hJRZxPkJNibpnzjCqfc2jloQVk1WAwiB1oXhkGJ0BNAtdIkiQ8TJpuDwh1tCFlz0Ss1WpR1MSio/R4Ze57sVio3+9Hpxy29FI1BmSmbPbs7CwKWfzz5MOpbyAk8PZnIC88PpAc41ar1aIkdm9vL+4TxUTwScHymhfDEPMSx1KlyFrkSTFQC2w81+WaGE/nUrzSkZAI8o61AE0Q+h0eHkafRS8Go/qT3ZdkmkCUKL6UbYCLgSTE8r9jRD3MYC4IW/guZCafGUD++bu0TaPf1hThn5D0h57//+9J+qf6EUbAY0Ep+4Ae0yMQxGgIAd4aZXRvzvtRWq8blxQ5cC8OajQawT5D2sHoexiC4Hhq0wtwaMfli95ut4MsgokmHAJWYgSAgtS+U53GT+4fL8a8lMvlqEEAVXDd5XKpWq0WxUV4Z+YFpWUNeB2hxnPdxImwhnhg+ASMta8daAIyDW6jWq2qXq/r+Pg4UrbIwt7eXkB31toRY6fTidSep9scnV1cXOjs7CzStGwjH4/HsW2cQh+MFsjU60GYM69dwdh5dZ+07Q+ZLwDykID/49SQxbyS+++3BQmkkn4lSZJU0n+dXrcSP0y3HYef6fq8wsxI7NwBvIGUDQ3iC3JxkjPveBH3MHihzWbbxAJozwSikMSoVIXV63Xt7OxkGFn3KCAGjxc9WyFtG6SwXwBPdFOp7Gg0Cn6B+BqoyvOgwCgT9096EIILaF4oFD5BkgFP8Wz0IPCwwY2bIyv+5nUSGBTuE+TDDkgqBb1s2bkXOAjWir8T15NF8BALlACLT8ggbetHMK5uVPk7ckZY1Wq1dHl5qW63G0rKezFIfigp8wKKYZ25NsjLZcFDWcICV3jkGkOGnCHz/HRDy3X938uMV2EE/s00TZ8kSXJH0j9JkuS7/sc0TdPnBkK51+PcgUKhkErbriwsFAuer9LzB8cQSNtUFt7UJ5VFcwF3a0suHpjsqMEZfhqA4iUQQmJvrk3OHJIO5afyrFAoRFEMCkC8vrOzI0nxd2l7CjHP6eQjzwKE9cYYKAahE9+VRy5ch+vn49/NZhMpSe/dAKeAYQDdMK9e8kzWBDIWpYbwwxi7waQdnBO3FB3xDIRLGDTSpaypl+jC/oPMhsNhGE/WejqdZghffpeUMSrInTsxlzuvUnUZzunBJxyfv8eve4MO3Q4kkKbpk+c/T5Ik+YeSvirpOHl+/kCSJPcknfyo66DAeCKvpmLxifd8YjAUDpscAgIZ3fvyPq5DT71SqRRVdsBQ/s71yuXrhpoIBtDbqxgLhW0VIKEBHAMZABhwPM5ms4ltrygpWQQgNJVyfC9kJqjGmXSgK1kL9sKjXMTdeGf2WxAP8+wYEr8+MBcugLMbUERXBIwEaAGOxdfN6zuYQ5qaUuXI+zESQPLNZhN7RJww5ii3arWqk5OT4HI8rKSW4MmTJ9rd3Q3UhjHC4+OUIHJBL8PhUDs7OxmOw7NCvqMVWeSeMZp8xkk+1pa5+DSOwHTwsyvsDeNlTyBqSiqk1weSNiX9W5L+C0n/u6R/T9Jfff7zf/tR13K22D21CzwT7FApX7Xnys64aQHcIFDMkoduvhNNUiADlMm9JZ5nvV4HnOe5XOGWy2XsVPTYGiLLc+alUinDTaBIvBfOAeNJCSzkF3MCOmLuPGvCMxEq4d1dUVy5fK7xxuwidEGVtkYWyIwhdBINXsMzFYRuNC0hJmf+fD1ZF5AOc8iuy2r1+oxFQhfmEkW+urrSyclJbNQiJXx1dRV7N5rNZlQQwmcQduAE2BPifIAXGuURraM4d0zOf/FeR72fx3hZJHAo6R8+v+GSpP8+TdP/I0mSX5f0PydJ8h9J+lDSn/pRF2ICpa1CoSC85ovnLDWLSV7fBVraZhecK/BiHwTm6upKZ2dnunv3bqahKRWH3CeQnh1qCDvPQL0/74XsosMQz+Q1CygdisC9EvtC8AHJCQkc1QBhvUMSEJ3ny8NLh5me43fPzPAmK+7t8+w29+skWT6V6yx5mqaxy9APF4E0ZHNOtVrN7Ngk3CDUQD7SNI1uRqQsMTh+RgBrQpalVCrpzTffDISDAWTX6MnJSRhf1hbj5qQusoo8e3WitHVinvd3p8fzYSBY7/x6cK2XHS9lBNI0/ZeSvnzD62eS/siPeS1JW0TgDLuU7aqKABAySNmUHAvgC+KhgXvwnZ2d8DikkOr1uh48eBDn6SGs5IAxDN6FmEWjCy337ApIdyCe170pHhHlpubc308hDOSbe048vlfrsd3ZdyXCmrMXA2MK5M3vxZCUMbwgI/7mG2Kcy/ENVhiNfIjG+jh3wffxN9+T71WKpEg9ZeyVmjyDP6+n+fy4NP5O85O33noryEIKqprNZmSGkE+u54aL7fCEZKzneDzOyLIbAjcGTni70XA04OjCdedFx62oGOQhvRrQK9zcQkrZ0lQmFHiPkAGdET4PB1hEzr4jHgNCUy5LrtrLcjEw1BCAWNigAylXKBSCcCoWi7E7DuXzDSvOBjv0zm84ATlI2dNrpS1R5TE/7DnPS9jB+5nvfG9B5gtm36sxIdqWy2WQon4Pq9UqauzzIRNeGaPphVXS9rRiEAunP7NbkA09pF3X6+s+Dd1u9xMFP8iEp5GZF77HD1plfk9PT9VqtfTgwYMIwajR2N3djTMnnCdxIlDaQvxP89xewJYnX93Z/bBQIE+Qv8y4FUZA2qZK8KwoMl6S/QI31QC4sfBF8XgLZWXxsOwYAK4HrGMXGpwBoQnpOITc4Sbn6UEiIWyDwUCnp6cRY5Jh8EpB0oTk2f015gR4zN9RTohFvs/ZfjdexNmQhFwfBh/Ug7fzoiBSqMViMaoWMSIYSxdI6i76/X6Ua7MueG6eDdQBCehEGwZH2qIhWpGz/hC3oJ1WqxVGAi8P2oD8dYMkKb5nPB7ryZMn0cBkvV7HCdMHBwfRmwEiFf7ACUeXUeYmn++/KRT4tOHyjqPLG4mXGbfGCOThkBNuTK4Te/5/n1wm9dMmNk3T6BKDEuNd8ZhUEt69ezcD/zmJFi6AWO/qatuNRlJ4XXYFnpycxI44ILgbAIg9QhdXfPdeKB6v4ZmJi70ICmPnA6PoeyiA3B7WgH4clvKPtCgK5mtEfUCpdH1mn59pQAWitC0J92pLbzPO+pEhYO++ZxHY8oziY8AphsLA0VMABIEBoviKOWI+SRu+//77+v2///er3W7HacLT6TRzVDrz46XZnvXIr6NzWT5/eeflvI1zL64nn/b7i4xbYwRQgPxAyRBASZ+o+HPl4W83xVgoU7fbDeXD8/k2U2nbQwAP5SkuroeAEwb4wvNZ9tRjTKRtl1kgLMrslYcw0A6NJcVmIhABqTCUkO8mV+4745hPEAdEIHNClWKhUAglwiuz3Ta/AQtCjtCGTIinWTF81ERwHxg/L4lmeGEWSI95ZB64d4wM/f1ms5kGg0HE5uVyOc4FaDQa0fjFDQ58B983mUx0fn4e5zIy5+VyWbu7uzo5OflELO8Oi+tDIHqaj7+54rv835QSzFca5hHvyxiCW2MEnDThdyYGiMVrHif56/zzyjZ/X6lUijPknODBixYKhSi4gZVfr9d69uxZsNK1Wi1zQGe73Q5hIsePRzw7O4vafycnOQAEdr/dbocA0NCD7rlsigHCE45AvDlZBSFZKpV0586d6PUHHPUsCkZPypZJj0ajQAaz2Sy8OQoNDPbzDHgWjFur1cqEcjx/Pv8tbZuyOjR3ErJUyrYXx0DyWTr6YrwxJvQboOvPZDLRvXv3dHFxEbwI5KqfMYEj8WwKxk1SGJJ8tgVZIYxCtpyY9b0an8b4f5qXdwPi49Ne/3HGrTECUjb/7w/v24E97gIeM7luEfMFGEmSxHZbWHbKg71azwtp5vN5BgpiefHqq9UqNh9xXDUQlt2AbrDyENozIPz96OgoTgtGoEn3ETNzH2w15jvYtefkU6/Xi4akCKxvlPJOSJ1OJ+PZUXxaj9NPASNEfQXKjgF2xcUzs70WyO5cD/ebJ7vgM4DXvmGHdeRcQIwd73Fuo1C43sV4cnKi8/PzTNUmjsYh/Gq1CsNAZqDX68W6wOlw7gKhgPNK3IN/B2GBFwAVCoXM1mpHwz4fvBcERkjpn31RNHBrjICn/ph8H/mKKmC/59jxNtLWQsI2I5h4Qz7vkMrRQ6VSiUo7FNR/ghyo4IMfoN30YDAID8c9uieQrtlyPP9gMIh0FIrvBVReow4ZBaPv4ZF39J1Op1qtVrHd2NEIoQTkp++H97iea7sn9owCVZHMNcaHtSR8QakoqkH4ySgg1DwH88lzE4cz0jQNNMYc+3Zs0ApGpFS67mHI1miMI2GOZ2ggIN35MG+kAVutVmzfRqa8OhIZ8aIu5FzaFrA5YZgntN0hfhpqcCfzouPWGIF8DIRnRwA8d+zeDqvq8Mrz9iiSbxl1MszhGwtCrD8ajTLFOoVCIZMdgCSiiWSj0YgOO156i3BKyuxFIHXote0IMQruNQ9O9vF/lIV7BrmAbE5PT+MeaMzBPM5mM7Xb7QhpyHHDlNO8k379PDceH+9PKTAZFDIINCfJ36vn0DFq0rbtPOjPDTxzDLlKHQXGnRZrTrw5+YpiYzBYW0IzFI17kbYZA9bZyVk3usgVxgHjhrNheJUiMo6R80zMTeTfD4P9/1pxAh4XStl4x3f98Xu+oIjJdMvNxHrNPvGaQ3J+B056Ks7rxml5naZp5JKBjN1uVzs7O3r//fc1HA4DPvNMeAk85NHRkUqlkg4ODsLIoWgcP+5IBLg+HA7jWh5HE0NjsCRFOyx4B0i+NE1DURFutsNKijMPPOVJvQSpTr7T02EoMChnZ2cn3g8yco6Ee1+vs6cKSdeG686dO5lmpTQxhTTlZCrWCm/sGQ5kyYt8JEUPwEqlEogOJwBJuFgsonmKlK2ORFZYg3K5nJkbR4HcM8QkhsDRgGe9uNf8Rq68zrxsZkC6RUZA+mSakAd0DgCldevpGQEgMyjBvTC/ezqHa0LysNh5r4cQ+sLX6/U4qrvX6+nw8DBgPSOftajVahqNRiEIxWJRx8fH6nQ6n4h5iXNRTrwU1YoImqQQevdqMNooB97ae/zRhATDSQsuV+TxeBwcCddylOU19KT+eD87M/HmoDYMAApDYxDibdbXvxMZWS6vj0lnHThrgOagIA8KuDy97KTxaDRSkiSxUxH0BWcAumLLNWcost+AZ0dGPcYnxGIucUReuu7OxqE+c8Pvjnilree/CTW8yLiVRoD/M9ybEhciiJ5y8rjRJ5lKLyA8C0PGAAtOfF6v16P1NUwwbDOeotlsajqdSpJ2d3e1v7+vSqUSB4x4hZo30Tg/Pw/v7Mo0Go2iQxB8xHA4jOcejUZxOAiwm4wE5B4oh9eBmvlWV2m6PSsARfWKQe+fyC7B9XqtbrebCcWoxSe16HUI7GDkO6StxweVwT1gHDFCKAwdnDCGPAthGdu1CUko80ZOOLOA+8PYe+bH04ukgN1xIFvsLGy1Wmq32zo7O4u1IFThGev1ugaDQaSJQV+eNWAu8qk+jImnGh3tvioewMetMgLSzcUQXkrrLK70yT3wwFeH4BCJeEEIHk21EQAAIABJREFUHrwNcJ7Ymh1n5JjJgU8mkygI8hOAEAw8HaEFisE9jsfjTK4cZcCr0+2GgUBMJpMQFjympDjim2fA2/uxYzwnJwGBFMizI4zMmaMVz2YQo5MNSNPrI8kvLi5iww3KdHBwEPfKfYCyyHB4A1fug/gZBMb7QDWOiPw8QHoierhB7cZkMgnjLinmCZmgGhQDKkn7+/thnLzRCLUYSZLorbfe0scffxwIDD4Fz85+FEIZ39Xp6WznAjx9yFw4Keiy5Kjh08KFzzpulRHIp0R4zVlXJopcfn4S8gQj8TdlrM66AkPxgsSHfB5vly/28f3rjUYj05qs0+loOBxm+AlKh/GwviFns9lWskmKqjUgZZqmIVDStmfh+fm5dnd3MxuJzs/PM+kjhAwEBKlF7h4IjGFjXmC66QzkbdjJ619dXcU1vW02Ckz9As+cD9XYmcfvICY4BVfS3d1dDYdDDQaDSF0Ss8OROPu+2WzivSA/SEMQhaQIedypgAIGg4HK5XKcSYjxATktl0vduXMn0oQcTgJ6wohwj6wFJKPXv+RTg64P/vOm115FnUDhR7/l/9/h8RGCg5IjLCiyd4rx93rMBJzE20Ek+RFZfNZzu1LWas9ms4DVlUol6tfpMX9+fh7W3rMMkvTs2bNoGCIpqvBQED6DUgKzISAxQNSh00df2papXl5exnl83W43FJhr84yU1MJbAPWdaIUVx2DQrJQMCN4PxeUeQE1AZIygd9rlGswzygmK4pn46ZkdMhwY0s1mkwk1kiSJVu6Qg5KiMWmj0dDBwYH29vZifkh9UrVJmOenF7OLkeeETCVzgvMA5vv9OylNWhVj6mlnT287ikUO8+NVkYLSLUMC0s3k4E1ZA7feCIRDLrewHv8D2XzBsNzE9HgPWnO7d0LwiEdhvzFQOzs7UTA0Go3ihF0npHzB8TRemgsrnaZpHL8N7E/T6w0/8AW01vasws7Ojt58802dnJyoVCplsih4H2CttDVADkf9b5B3np7c3d3VxcWF9vb2Mm3LMJKSQsmI/WnljQIRRxOLuzdkvtrttsrlsvr9fiALDAdzxXxDvHHICRkH5p+NQ9wTaT04BW8IO5lMMvsVQFx8hhRwr9dTqVTSxx9/HL0eWHuelbVl3wVIzJ2My7aTjBho141XPV7YCCRJ8iVdny3A+IKk/0xSV9Kfl9R//vpfTtP0H3/Ga4ZyMjlYUYSCgWCDBPIkiregBooB/fFgEDdOLrqhoAxXUmxjRSg5w84ruOABOp2Ojo6OdHp6GlWI3LO0ZcQdyuMxyXlT2kzabL1eh7fkmfk8RhLoTqz94MGDmBcv36Vgh89K2UyJpEyL7G63G/H4er3OFEhhQH3fvyt3u93WcDgMBYDkY95AS/V6PTYR4VUd+h4cHOjJkyeBKFjHSqUS7d5o9cU+iDS9LijC4JBt8J4IHMUG6pCU4RYuLy/jYFYIYT+qTFIGtUEgegqauhScESGTZwTyBsGVHWPg8+GcwsuOFzYCaZp+T9IvSFKSJEVJTyT9Q0n/gaS/kabpX3uBa974DyV1WORK7wLuOXKvImPSvPEGi0XsxrVQwkqlEsQS+fHxeKxWq6Xd3V2dnp4GWQZUXywW0Y8ARXEIicfDqHGPxKvE40BuTt/lMzwbqczxeKxerxckGMYAA4nnAlUkSRLknPMpjrQgIr3IBv6ElBwn8CK8Xip7cXERpwNR1uy5frww3wtxhtHn9dFolEm3rlYrHR0dRaEPFX+EB91uN8IVFK3RaKhcLkexk6RYm4ODA3U6ndiERWUjDoYwC+PoJeWgQ36/d++eJAUvg2FCRuF9yLI8efIkI+P5mgAcgzsy15NXwQUwXlU48Eck/SBN0w9f1c05fAXaOZPsaRO3qs4XsFAog9cWIOCOPBASBADvj8ADOz0TQH08cf6HH36o1WqlBw8e6PLyUkdHR7FLj9SUGzevYgO9kC3AuyCMV1dXGU/EZiaQDXX6fMZTdjDsnU4nEJDn6FEW3s9cwCM4eYp3lRS5ePZWIOTejMN5Fw9ZvPaArdiDwUDn5+eREfLy3fV6HR5+NBppNBrFfFxdXUX13xe/+EW1Wq2Iu+kfCE9AhSNhCjE83Zjgf5iLNN2e2fj06VPduXMnSEsMX71e18OHDzPHrYFikVFQnR+Qy/zhkDzblU8T5vXjVenaqzICf1rS/2C//4UkSf6cpN+Q9BfTH3IEWX445PEYFgFz0k/ahgX+d97jtQSQa3luAS+CoNO4A6JoOBxGGWi9XtedO3dUqVSiPPfw8DDe9/jxY52fn0focXh4mCHPyJkTOxMKSNv97MViUb1eLzr3zOfzOBCT04TJZS8Wi8we91KpFGcJjsfjgN2FQkGdTicUjxORSB1eXV3FSUzEuyAm/uEJMRYu1KTXgNegDMIFFJhCHrw0/AFGxwutUHiIwEKhoPv37+vjjz+O1CdKzGYeavknk4neeOMNbTbX+yOQj0qlEkeQQVxuNptAUMgLYQtzQyZjMBio3W7r+Pg49h4wVyC1w8NDnZ2dZdBrmqZRSQjabDabGg6HMRdeIOQKnq8jcNl/VUbgpbMDSZJUJH1D0v/y/KW/JemLug4VjiT99U/53C8lSfIbSZL8hr9+UziAEEJOocxeS0/cBNML9Efx8CwsPPvsQQ1AfkpJgeEI/s7Oju7fvx8CRAUZmYanT5/q/Pw8hD5Jrg8f/eIXv6iDg4N4Fk4jhtyC0caz+jHZKA5GhOo+jA1nFX788cd6+vRppL+KxaL29vbU6/VCaT3vzFwB76mD4F+xWIx5ZBMVGQOyK2zG8lw7dfSc9pxvP+YFXBgNeIfZbBYnAIEOmBcQ3J07d/TOO+/E56VrpMB6UDTlUJ7iLV5jizaNZdgOjlf3vhH37t2LE42ovTg/P9e3v/1tvf/++5K2hWyEYPfu3csoMkaC9U7TNL6XGghHC670HhrkCcFXSRC+CiTwxyV9M03TY0nipyQlSfJ3JP2jmz6U2uEjiR1Okn9g0ilAd7ewXkXorLZPZKFQCM9Oqol+9sTcHpN591qvcUdxCSUoGy4UCvrBD36g09PTuDYVc8Sm9MCbTqcRYvgWXrILIAygPzEqKMQP/wBy4q3IvU8mkzhjj+dkvugZgBGQtoU5GJl2ux39DFBsYlyE1w3C5eVl7JEAmksKb+3NPzy8k7apUmA694FhpviJ95OV6PV6UajFKUEcGsNnC4WCDg8PIy4fjUaxuQfeB86A+QFZcC++nhcXFxoMBhoMBioUCnr8+LF6vZ5+6qd+Sv3+NQdOmPPOO+/ot3/7t4O8JRUNGkySRLu7u/E5kICHtxgMjIkXFTlafhXjVRiBPyMLBZLnh448//VPSvrtl/0ClA5hwJs9/76YMCdSfGMNk0fBB54ESCopvC714Xn29v79+1GcBASvVqv68MMPo2CEuBmjxM9C4XoDjzfHgEl3go1jyAllqLJDybvdbmz6mc1m2t3djV56eNPlcql+v69Op6O9vb1QIrw8c0MsWywW42+gCIpsIPPW63WmNwLVcSgv/6dtOB5uMBhECAP68qwO8bYz524gMeoYEwxZr9eL+gDqHKRtzf9qtYrUKwaKHhLsHCQzQTbF+zdS8uwOAEUGYRaLRX3nO9/RnTt3dP/+ffX7/fiObrervb29OAoOHoQKUknBEYHykEmeJU/+eUYg//+XHS8VDiTXB478UUn/wF7+L5Mk+XaSJN+S9Icl/ac/7nXzaMA3WeTTKk6A8V4mEOiJojtB6Kk0vhMl3mw20VVWUlTmLZdLTSaTIJaoYmOhQQsI92q1igYa3A/7DxigCfbbp8/rAIg51+u19vf3Q2i8is3f12q11Ov1wiBeXFzo9PQ0TjhGoIHBOzs7mXAKJCUprochIutBw1SafdJGez6fx9kNZB8kRRqVkASCzbsjgQQIWbgPjBOcACnTdruthw8fan9/P7w/uXpJwX94ubcTi7PZTM+ePYu9GPR/9E1VhJ9UFHJUPAiPNOdyudSv/dqvab1ea29vL0K32Wymn/mZn4k1B22RpSLExUD71mOP/538di7rJh15mfGy5w7MJO3lXvuzL3NNV2weEg+BkKD0Dp/88/msAvAPbwcaYHjIgEeGwSX1BKzcbDa6e/eu0jSNXgDefBKCiDiW2BSCipy5dA0ffWuzF9LALtMOi/uhXwEQHUWCGJQU3wWchxGnExHeFsUEablRwAD5TknuGSXGKO3v74dx5d7X67XOzs7C0ztf4YQv3w9Udi7Iy8QxYuxVwPti2On9iKMol8vBzsMDeKrNIfZqdX0qcrvdji5MzCNNUTAChFMXFxfROu7dd9/VH/gDf0CNRiM2E1GOTRs2r02hnqPb7UYYyX3gIJzrys8XOvCqDMGtqxiUbu6x7kbBJ4LhhsH/7taXfxgEBI+8r7fnRonu378fSupQ9d133830v2eDDl7EW3B7eSqVh5eXl1GWStszLyNFIYHqwFQnypwDccbc54ayV8g9TgtmUDiEF5cUpcU8y/379wP6TyaTUAQIVG8Jvtlsom4Chl9SJhQDavu6OtfjYQDcCHUTsOo8L1yKx98oHzJB8RFz5mlh1oZwhDS0F5h5eAehR6hxeXmpH/zgB/qpn/opPXjwQN///vejmcubb76pfr8fBtOPj6OQCh4Ih+NhpCNcR8Lu9F7FuHVGwJXVB5MhbVsys5heXcY1UIx8elHKTqjXo8MwQ4Z1Oh3t7+9rNBoFUdXtdvXhhx/q6uoq9smzyHgrSm1ReG9kiQGAsUaJWHDfKstzUpkGkblarTLbkzFYxKSQXPn0KrAThcaAkEcHebhikrVA4UAPKLo3KYEPAPZL2TZv3sLM18Lz6cyDV416eTej0Whob28v7hc5SJ7XWWDIyHJAZiITTvwC1TkklhCN/hE0WHEehUKlk5MTLRYL/fN//s/1jW98Iwi/NL1uOtNut4MDqlQqmf0lbD7zZi5wSL7fg+91pPMqjcGtMwJS9mAGHtw9u3uNfOGQx7iOHIDXbjBQDpBAobA99lqS3njjjfBEVMqVSiU9ffpU1WpVp6en4S3L5XKkHb2Ckb3sfB/f6Z4fxYE4y0NnctaSIm4kM4CHc4FO0225LPE2sSdKAzHmULRYLEaKFJIvbyjZQcn3k0JjcwzX8gpAlM43Enn865uCWFsMA30P6vV69Exst9u6vLxUr9eLw0ZYA4+5CbX8HrzojHvgvqkqpFCr2+1m9kvwTKT8KpWKOp2OxuOxptOpBoNBlDfjhN5++2195zvfCdnGOYBgWq2WTk9PMyS1I1jW1MNk58FehSG4lUaAkX/om4bH/84TuAFhUlFuzyygRBBRCEan01Gv14ttwfv7+9rd3dVv/dZvZerf0zSNWJItt8TpKL3zDRgs4CcVe7DDXIP7w8g5q035LylEEATP7Z4Hz+fGFA/j+WhSa45q4Doou8WLEz4g0PQTwGvzbKQLQSCkLL06zo24l0V7dSWoglbmeGwQE+EdyuXPx/1imOEnvIYhSZIw9ryf7A1oj0IywqTpdBoEMXP+wQcf6Gtf+5q++93vhsHtdruBcpgb5gkiGSOLcXaSljnieTw7cCuIwc975EkQaXvWuwu2pIwXYTCpLrw3GRVCARSxWCzq0aNH0XK8Xq9rf38/DrTwBhWQRHhs320I4YTnQfn5HJ4JOMr/aWEFfMRT4RGB6YQGkuK+aJkFhPaKSiCxexpCKhc85lPaNmtxAcazomSEFl5+7MbYsyOsByXKpP0IO9xbo1yEO3yeUAeiFATlMf94PM7sAQBp8ZyEXYRDrvggin6/H7tK2WXIMW4gm/V6HRvNaBu3t7eno6MjDYdDHR4eam9vT2dnZ5FxQG7hcFqtVhRhuVzepOhuAP61DAec8cx7dN/vLmW7DEnZCfP3cQ2EOW8snJFFYNrttrrdrk5OTjJCOhgMIp5GOD3FA1zlXlAGUo3Epq4EbvHxhKTyeA68bqvVCuX2XZDSdhceCuMhFOlL37Tjexhgs5lLwhRPlYEMKHqizp44G8WFdMTQ8FnWgdgexOEpU+aF33kf64sCOTcBuQt7TxhBqrBarUZ1pTsPjAL3BTrBOEIaUjaMMWs0GmHsIX/hVGq1mubzue7du6enT5/GeQvdblf9fj8QAHK7Wl23buPvTk66HrDG+RDAuZuXQQW3ygj8sOHQESHxyXF4i/IwQQgk75W25w6yEMTVkvT2228H7CwWi3r48GHsrBsOh3FKrlfBkf/m+gi7W26E2A0RsBfWHmVz5IP3hBH3zTWUMeOpvSTZOQ/P/TOPbmQQfBSBuUJwJWXSprPZLBTYC2rcAMOyS9kTe1Eo5oEQCBIPg4Iy+wlFPDvcCwYWLy9lz7DAuOFxIWXH43EGGRC+cF8YDEKt6XQaslQulzOGAS/u6Iq9GXQgeu+998JgMVgbDscdDAax1i6reSd1k6N8mXGrOgvlGX4fDu34u8d8biBuIgVZoLwRQCnwjLVaTXt7e8H6Uv11dnams7OzKLyRFEpAgQi5Y+LefKchr85jF5vHfNIWikrbRceYgAAwKC50tPf2VB9FQeynbzabarfbkcr0MAQ0g8J5So9iKIql+Dz7J7hXjKJzEd5Aww2i8zLMAQoI8pG2JbQU/0gKY+f7Scg+YGycG8C4eyMYirqYc4wFn/NuQhi4fr8flYEUamE0PTOSJNd7COBtkDEPpQhHkEGauPLMyM1NuvCqx61FAnnrd1Mcn1doJwP5LH+DGGQxgJXOFBcKBe3v7wfhVi6XdXBwIOk61z8ajTI732gsQb5f2raHJqWY97ye1sTrucGijJT3oUzr9TrDL1DaXCgUop8AUBi2XVIYBTYGkeOmbsC7ArlR5F58fngmFAskgmGStl2geT7WqFwuR6ESNfrMP3E8YYV7Z0dx7P/gnplPEATf6RzDZrMJBEMYB/pzUtazMYRtjiip88BAcooRn8dI0tyk3W7r/Pw8NkRRqJYn/tilSf+F2Wz2CRTrMv+qawSkW2wE8mnCPEnoHiRPaknZw0qAhVh+aWtcHAoXCtebTmgNRUZgMplEpR/wbb1e6/j4WMPhMAgiyD/3osTceF7idaAtUBivQazIHKBUbtgwDBcXF7GTkd18cBncB8rF7sUkScJjORlG0ws3uHhZvtfrEvi8IzEMCcaD+SJlyPdQLuvryr1K26Yf3LvH7DwDYRHblQuFQmRyPFNCZqBer+v4+Di4CIwWXYp4DoqH/BlZD74T+SgUCup2u1qtVtFheTgcRkfm3d3dKLGmHRuG2PP/ZIXo4MT35TM6n0dmQLrFRuAmA+CGQNIPtZL87oaA9yNsXItJJT6fzWZqNpva29tTqVSKpqLkfqXtOQBkBhxauwKRWpK2eyDSNM0c7onAgxBIZbLQGAu65vBeYKy0bVohKTIdDnOB2BgQCmC8br1SqcSuPLwqRpZrOBnr5dcYAryopOiNyN9Rvjxa4+/eBcpDFH/NsysYB+ZNyp7LyI5AipkILdhIhdJ7dSZt0dk74nUZToDOZrOoF9nf3w8jQMgE9O/1ejo5OYnDb1nffFFWoVCIlCP8w6c5Qg93X8W4tUbAx6dxBa7oHj85vPZY0wWE4cQhUJMqsULhuniILaJszz0/P9fx8XHEoMBSFpMtsEBeBBmPwz1Ceknbhpk3wWTe68SYx8FwDcTvXvyEcHtnIHgMymDxdsBkQgHfB+8GkOFGif6A7tHx2FK2FkHapiA9bveYmBJryDvPDmw2myhHvrq6ikIqSqThK+gSRHs2CsEgSrk2iM1DGgqQ0jSNMMxDk0KhoPPzc5VKJX31q19Vs9mMNCBFSnAQ1B84OnUDylrRw5B5dOLbZf5VG4JbawRuesib2FFPj3h6xz0NsS5FGbzuRmGzuT7Gm2IcPAVHW5XLZT148EBHR0c6Pj7WaDTKeFa+g5jboS0Lj6JK2Q0inkpkIwwxOOwz1XMonXvqJEkyW5UxTDyfGwW8mZfgSltyjGdAAb3OgvCE+3NPRWYDowPn4sVKbGDyNQQdJcm2hRqbdoDMGBZPy3rtgj8je/RpV9bv9zO8hGcdMAT5FmhpmsaJzZVKJcrGvWaC+5tOpzo7O9OdO3fiODSehXqGTqeTaS+P43JEh5NqNpuaTCbBMbl8f17j1hoBV+abQoGbhkMrv4b/3dlrhiMB4DtC7T3pW62Wvv/972s4HIZAuIEBPfBd0ra5BooLQYQQsTHm4uIiWmXD8nvj0s1mo93d3UAPIA0gMd11HPLu7OyoXq8HaUgKDKGXtturIetIfdImzZEGKVCUkA0x3sYM7iNN08iAzGazEPL8rkdgPoaO+XbDzlpimJ3s5T2EG6VSSaenpxqPx+r3+1FtSHcosi+OPkAZjtz4na5D8EKsO05kZ2dHp6en+uijj/To0aMgHyVl9irQkdiV28lTRwNe83ETCviJ4QQYDhEdGnvRh6cN+elhAJAU4cP68l5SW3hTJ738vTQRhUwkJ+89/vAkKJYXrwBzUe7VahVttWDr8fx7e3vBHdDMlP31nl/HCyLIfAf58F6vp4ODg4jn8axe1oxRZP44ORdG3wky51Z8k87V1VUIMIpFGo+qRwSf+ZCUMYhwAK4QfJ9nUhxVOaMP6cm9eNdnDoDlNUg+5lzaGhw3DMgd3Z1JAUvXh9DSWo39Al/4whdijheLRdQLJMn1GQU0gmVOvbZFUoQy+TAqj3Bf5fhMRiBJkr8r6d+RdJKm6c8+f21X1+cOPJL0gaQ/labpILmWkL8p6d+WdCHp30/T9Js/zk3lvT3C6TEynsEVX8pWDuKB8/35nfHGWOBJkiSJHD+Me6FQ0AcffKCjoyMVi0UdHBwErPXecU6YwQI7CUmTkclkktnRR0oPQwKcJJdNmou+h0BrLxmmcMavm6ZptP1muypK6J6E72VuIOmox5cUFXd0G8JYuceuVCqx8Yg5IAPD/gLeyzPCRUAo+jw630Oo5zUOTipiNFqtVmRv4CpQHOL7zWYT/QPIDng9grQN4zabTcT4kJySAmF5qfjR0ZHa7Xa0NcM4Sordid1uN7ILnh1w4wpqc87AZTz//5cdn7VY6L+V9Mdyr/2ypF9N0/QdSb/6/HfpuufgO8///ZKuG4/+WOMm8uOm2MjjKiAenoKJ83gdSIpH9Hj13r174cnxFsC94XCoZ8+ehWB0u109fPhQd+/ejXQXZaTcv6Qg1+bzuU5PT3V6eqrBYBDexONslAjCCIgPedftdgNxgA6Gw2G0+pa2DHyr1QoExO5GevIBVYHTaZrGZiM/bJR9AO12OzgSaWtkvYqOLbGc1uT8AmiJMIR1pAGn8zqsD5kO5g5jwfrkuQ36BlKERWUk90lIBRE4nU715MkTPX78WM+ePQtjxvslZTIB/J98/8OHD3Xv3r2A8yC2NE31+PFjzWaz4JdoMTaZTFStVnV4eBhOyRXfnQCIypHJTSThqxqfCQmkafp/JUnyKPfyn5D0h57//+9J+qeS/tLz1/9+en3X/0+SJN0k23fwMw88PnGYV7kB2TwskLLFRc7YwwFgNLzVdaVSCfgNCQh8WywWevz4cSh4qVQKJWNxMQ7O4ksKD4zXxjO4Z8NbJUkSsaC03XHmVYe0A/cSVngCyCw/yWez2R6+QSbBY00PDS4vL8MIrNfrgNte1kyLc5Sj1Wplro8xdiUFcnPvvo5sLWY9CGWYG+A+mQkPQ6gP8I080vbcQa+14OxFj799vwcH1roR9sIu5gQP3e12A/HQ64Hrzudzfe9739Mf/IN/MI6Qo0UZ+w663W6m94PzWMgG3ZpBMo54X2VmQHo5TuDQFPuZpMPn/38g6bG97+Pnr/3YRoCHZyHcckqfnBCUT8pCfc/x43ncq3A+AN6JfeLNZlP/4l/8Cw0GAxWLxfCIQGq+h80jxJoe67OIbpgQds8FE/eCDEqlUtSkS9d1CZCHdLWhsUfeAJXLZbXb7cxBIx6DMhcoxXp93fOfXDaoisF2YUIc5og9/ggrhox4mJ8YAzx/fsvyTTEvcwxiw2h4VWStVstU5GGM6Ji8Wq2il2Kabk8IJlTAQIEoms2m6vV6HDXHnEL6gjQ8Zcq6ewiGIfi5n/u54DwwRvV6Xbu7uxqPx5mTnrwylGfGUeU3Fr1qXuCVEINpmqaJtQ3/LCNJkl/Sdbjwo64dSo6F9pLLfMYA5ec1GGMWjcXCI7EoviusVCppf39fT5480enpqdL0upNOr9eLcAEPTRkt8G25XOrp06c6PT2NFBse1TkHaVvfzuu8j2ISj937/X6gGVd87kG69pB0L67Vatrf3w+vjIBBlEGeOZ8BYel7+Zlvh+MYC1Kaq9V1d1/35NyPe1UMCHPsc0A9BmvO+nn1HgYLInc+n2cIYLgMwgqIyVarpXq9Hl6bjAxrUSwWY75Zm3a7nam25P0YWRSR5q6Ei3zv/8feu8ZItl7nee+urup7dVVXVV/mdshzJUVShAIaAaEgThAFkGIkMGQYSvwjusQ/IkD5EYBQEiYGEjkwQCCxA0OCBCiQ4VhwFAtQYEuCf8QiEIASRAphaOqIOreZM2eu3V19q6qurr5V1c6P6mftd+/pQw6nm9RpMRsYzEx19d7f/r51ede71re+/f19vf/++/r4xz+eSysC970DEfPsyNXDw2LG5Kqvy2wg2kqS5IY0aTMuqX3++RNJd+x7t88/y11pmv56mqZ/JU3Tv/LtHlKEPg73PR3nhJ9XeAHhXOBRmqmpKb3yyiuanp6cOUh83Wq1dHBwoDfffDP+v7a2FnCPAh+QRL1ej/LRzc3NIP7YWUjcPRgMIjZ3xp2sBTn6JJl07WFrMu8BxMWjQLqhfDxHmnRILr47l5cue0zuxT6e7gPKQ3IRC7Oxxll0EIaUFQTxbNaEEAQCFgPtey2YF96PccKJTE1N5dJtGDk4Cbx+MaVHaAPhOz8/H6ESMrC7u6scgetSAAAgAElEQVSnT5+q0+mEA2AeMUA8jyxIq9UKA0SIsrm5qZ2dnSg8wogdHR2pVqtFcxjnHjy0Km7x9pT5VV6XMQK/K+lnzv/9M5L+hX3+08nk+ryk7ovwAVI+L+ppJYTKhdLJE/d+CCKLh5dzQovF5my7hYUFffDBB9G+i5JQL7JJz1OLVKZtbW3p8ePHkTLjyDKaSaZpGuWooA68HyW+cA1UuHkjUzzSeDwOgcTwMAd4J7gBzzYAzX0TDIQYwu0eDg7CGeqjo6PcFlw8MeW5rBccCH97cYyvK0rIzj52CboB5994UOaB6kZJud2FzoXAwxACMGccHVcMLyXl0qtsJ2fjWFE5SScfHR3FIbWuqNvb23rzzTd1eHio1157LTZ9gQjW19fDgDEfyDHOqkhofy/QwPOmCH9LExKwlSTJY0n/vaQvSfrtJEn+tqQHkn7q/Ov/UpP04F1NUoQ/990MiPjdYb97KybZyS3/jsNDFN49P95PklZWVnKE2czMjF5++WV1u109fPhQ5XI5yEHYalea+fl5jUYjPX78WLu7u7l4E6UkriuXy7mjz/gerDYEIcLKEVzl8uR8Qb4L5HQvRHETXuS1116L2JX5OT09jbSTZ1G4D/O2sLAQRoMOR/ATnNYERIUVRxkguxYXF6NfH8rsz2SNJeWUy7fzooRe5ES4BYFLaAO3UK1WY5MO7/Lo0aMczHbD4lkkb9ACwhqNRtFundQkigqioRLw4OAgzj/Y3d0NVNlut/Xmm2/qJ37iJzQ7O6tutxthGecjEOLwHN+PQfEWtQk4wqu8njc78Lc+5Ec/dsF3U0m/8N0OBGuIgheVm3/7dzwz4KnBInnokBDCR5qcAkNqp1QqaWVlJcqCJQVX4LsBIaQWFxc1Ho/1/vvvq9frhXGias8PquT+dAZCkPguHWnca2M8EEa8IJth5ubm4gBOL4p6/fXXA+JC1kE2eRhAmIERAq3wMydTOTgT9MOJSxyWSgjiHth3Tjqx5SGCKycywOlFc3NzYRwZNyhgNBrFIbAoCBWM8DBnZ2e5PD6FT4eHh2G8payAjENS2GLsEB2FpkEI4QpnJjK2wWAQKcqjoyPt7+9H78F2u61XX31Vb7/9tvr9vg4PD7W4uKjbt2/r0aNHuboP1pTQyJHt9+L6yDQV+bBYh8+dG/DvXZQxcCbeGz0wkaVSSc1mU0tLS3G6jCR98pOf1OHhoe7evZvb7gqSoAIMEuyDDz4IGIxycZSYKxldbZaXl+PEIBSP2J2KNNALC+4Hh0DGoUgUr1Bks7y8HKlFPDwknHtT5oCzDd1wuvGFQYc4Yw3w/JQSU1oMNGbd2OEHrOVifngG3g4kwHgl5b7nxVUw/8vLy2o0GqpWq2o2m6rVamo0GqHokMDMCd7dMyZkFDDSxVCG1OlgMNDm5qYkhUEijHGUSnYGw5umqf7oj/4omrDAs4zH4xzSxCBiQEGUTnL+hYUD3++rSATCBzghWMwC8F0uvAqev3hPOr+QSnrllVd0cHCgdrsdzSK8T/54PNbi4qIajYZOTk70p3/6pwGVqSvA0JC/Jt8LiVck0IDE3tUHCI0AkI3o9Xq5TTPsC0iSJPLcoBMnGz0lJymHohYXF8PYTE1NxfZijIoXMwG7nd2nmw5EGN603+9HKME+B/f4ICDeE6QEFKapC6EXYVK5XA7U5XUe1DDQ8ixNUzUajVDM/f39eHfen6zA0tKS0jSNnYgQijgA0r0UXZ2dnandbuvWrVsxdnZTIkusc7PZ1MOHD+Okou3tba2vr6vdbsd3cA4+RubJuxthwL8XvMBHxggAP/0lPVd8kRX0CXNv5hYVTwMpxURSHEQsyYGUwEbSVXjT+fl5raysqN/v61vf+lbEjQgMRBoCTRsvrmI6T5p4cnLuXLDhGBeUgZJhahXckzcaDa2trQX8xZhghDxb4ulJPCXK6qw6zwBOE6djvPD4vBPpNU/FDofDmAPG6zl18u/MM8QkhTIYHObZT2X2yj4PpUqlUvRcXFxcVLPZjB19NPkAvVEf4eQnxsufy7zBPfA7P/RDP6RyuRyFSKAUdh3Ozs5qZWUliN67d+/q85//vGq1mtrttk5OTlSr1XTnzp2I+aXsVCgPD7yW4KoNwUfGCPz4j/94xIPD4VB/9md/pocPH0p6NiTwQh1PDxYNgudcCQ9QulKpFF2BarVa1HXjSagGRMk4ZOJrX/tasMMIPN6Zxh+0mvJaePfudNdBgDxFiNECMkIqulf0KsKVlRXduHEjVysAnGbHoiMTetkhzMwvGQtguacwHbU4x+GHulJpiZJhGIib/SCTIgrAmPMzD2d4Pk7C42Upq/EnPeqKgqGmSzKcEAhRUiAXtu9CRvL+cB5wLKzRYDDQ22+/rVdffTWKy+gzgQFGtqanp7Wzs6ONjQ2Nx2OtrKxoc3MzkAVZDt6RUARD7RkJL4i7qusjYwR+//d/P+JTvMYv//Iv64tf/GLue0V+oHgVYyjPecOu12o1SQr4CDkGtIYnwCNKE2H75je/GV672HGGGBFP6kVKCLSU72NAgQoIhn4BlUolClswBnS9odU3MPLWrVsR1gDpUUDIKsYBohmNRur1erGfXsraoXMfagCcrByNRlFJyfzgLaneIy0HCgFRsTWaeWE9+Blryzzxnt60BLTCWEA5KE+z2QxYDz8DdyFlnZvhNaTsBGI/UIaLZ1HxR/hFDUiapnrw4IGWl5fVbDZj7nwOkDdqP8bjcWxc8t2XU1NTUcKM8yiGt96/wis6L3t9ZIwA+W9evtls6hvf+EYYAVf8opfH02PpSSu5IGNhp6am1Gq1olCFWgHaShEDEneTrvvyl78cBgYBTtM06gq8Ao36hWKREqjFBQAlR3HJ/ZMxqFarUcnH+wyHw8gADIfDKHN14g9ijmfgLfE8q6ur6vV6EdtT9IMXZKz+N4aSXDzIjfuiSE4E0o2YmgJgNh6duZKyXaIgAAQdT0woBAvvBU9udMfj7EgxDD8GiPWAD6EMmxJgHJCfccjnThKD9u7fv692u62Pf/zjqtVqz8T07BxdWlrScDhUu91Ws9nUwsJCIJBarRZhjHeaAgW5zCHrXFcRGnxkjAAXk82L8n+/gEiu/A4Bi2lCL/mUFGWvxKReBgsBh1ItLy/rK1/5SuSmJQXjTztySCOe5TwEiipllXb0vcfDejnxkydPwiMRulDU8/TpUy0tLeljH/tYeBkadiBw7EAbjSZdkRAgwg1gbpIkkWYEmWBAZ2dnIz/u8+rsfK1WCw4AfgBPjseU8g1U3Thh8Pnc+QQ4BTIa/B+DggLxrNFoFIin0Wjo6OhIc3NzUQHYbDb15MmTXLdjvDaICU97enoazWUgbOF9KDBi7qkT2Nvb0/b2tj71qU9pfX1d+/v70a2atT89PQ3kxTOZd9DCo0eP4v6ETk6KY4ipJGXOL3t9ZIwAnojr6OhIv/IrvxLw0FN/F5EjXulGLOjfx+uQXut0OhFPlkol7ezs5DaKDAYDvfrqq/r617+eyxQAtyETDw4OwogUBRjFgVDyNBWVabSYrlar2t/fz2USpKxrMsaiXq/HRheUzvsT0BSVq9lsSpoYJ87SYy69ZsAbknozkF6vF5tqaF82OzurTqcTcJ+5Lc4RikYxk7PueEyMMbAbg0BogdEkRcq9er2e1tbW4jutViveDa6F8Ik6AWozQEg8e3FxMdYFI0kFpRctSYrCMoqkMEBJkuju3buxRru7u4E+QJee4gbJHR4eBqfAfGIckWcunIs7uKu4PjJGoFwu6yd/8ic1MzOjX/3VX9WtW7eeERqP/4gLPf/vqTognsMqDIWkyGHzHAitZrOpra0tvfrqq3rw4IE2NjZybDbwkPJeL+6RFMQgMNlzxfQAGI1G0c0WuM0fYkn+nSRJrtAJSItgSxmbLClgZb/fj3deXFyMf/tmJYwgoRMpOsqkNzc34/vHx8exHRbysdPpBEkGamFXHeFBvV5Xr9eLPf8gHs9g+OYYGoOCxvw0Zkmx92I8HmtrayuIv3feeSfSphhpsgkUWbEHgjUkNJMychQykLmgeQychIelu7u7mp+fjxCu3W6rXC7r05/+tD72sY8F+vBsi3dHZk09ferOzR2Y1wxcNTn4kTECSZLo937v9yRJf/AHf6AnT57o1q1bYbGLcVAREfhkFRliPA1pQJRgbW0tvDLbUcfjsV5++WV1Oh39+Z//eRgaBAaoCExj6yx5dGr2EVZIHPoA0KUG6F4qlXIKBopBOIDB3jgDz423xQBSSstzqtVqbpOOlOXKEUJPtdHNyFNqdFPynggea5dKpRw/QSEUawLqwGA5CuE+zA9FQBTpwJVgADBYGFbel3FPTU0FwpuamlK329XS0lKOr+GezAUG0GWlXq9rcXEx9gzQ/HU4HEZoRu0AsJy05O7urt5++229/vrrWllZ0d27d+N7EIxHR0cRWiB/kINeEs78uSP5XtQLfGSMQBHaj0YjfelLX9Iv/uIvBhIoxkcIlJODwE//DkYBVpj68tXVVZXL5VjsXq+n5eVlSdIf/uEfajAYRN660Wio1WqFcBLLkRoE8rHACNre3p5OTk6i3RQLD4FH2o5ctBOHIBgpa4YCDAWqo0iScsQmyuQdgCRFLM678X4YSAwExhfDiLHAc0IAkmGZn58PIwa7TQ0Ayjsej2NsIDXWiBDAQzYUExTl9f2Qg4RSPJOwh2Yv3W43UAdKDsIifINwpb4DNOdGEmQxGAyitThjgAhFJjY3NzU7O6vPfvazevnll/X48eMYGwoOp7S7uxuGAOeAwfbMlrdAI3S7qgzBR8YIQKxJUq/X02c/+1n98R//sb7whS9IyjPHeBJYZC8kcmKFzzzfzP/r9boqlYr29vYCyuLlaLCBINy8eTNn/YkRi3XyGJqDg4MQaJRRylJOxN4sMujBawqAsngunpOmaSg/UJ0tymzPZastxBWpw263mytzBf6enZ2p0WhEAw4nrhqNhvr9figcxgTEQE8/mGvWAINWLk9atIG2SB0i+BgAvDdjgqT0o998Iw1NO4HuXmHJnCdJom63G8/c3t6OtCHZHww38wIPQEbGy6gxRLVaLVK4fA5Hw3M3NjZULpf18ssvh3NqNpvRqn5tbU2Li4vRMxJUwnp5utT/LWWl5X/pjMCv/VrWipA00G/+5m9+6PeLaULf3kns7AQXSsHvtlqtMDzE18vLyxoMBnrrrbc0PT0dgsGziltrCQGIMYfD7GxCPB5KjuBT5oqCcJEWpMgEz0HogfDxTgiox7JpmvXuI5RgLoD7bL7xCjtpEt+Suur1epKyc/xmZma0s7MTnhvhQ7n4G1TU6/VioxPdc/DUGFrCFDyah1cYQ6/hZxsuikM6lbmAs/Edjh4qEgKidDwbHgO+ByVmbkBIw+EwEBjFZZT/whn4fPb7fT18+FBzc3O6deuWut1upAIxvGdnkz6O29vbQUY6H4AuONJFLq6SHPzIGAEvCgI2/tIv/VLuO04OeviAYHp4wHc83cci47VgnqkUHA6H2tnZiTjzjTfeiN1gvosLeAw/wI66zc3NIIKAdUdHRzo6OlKj0dDc3FycYuR1C8S8x8fHkZlg8T1eReEwZqAgjIlXEmL4aMkFV4Ags4EJjwLjTVqRzTQIebPZjDkGWSVJEj0O+Jz0HMU6rtgoNPv1EWbiXYwm7+hZEQ9R3FOynqAE5suLlTAKnAeIAaDeP01T7ezs5FqIYdSQRaowYfSlyVZ0DBCZCcIZ6k7eeecdfeYzn9Ha2lrsSuXIe+ZuaWkpkAXv5tkw3tUNw19KI/CdoI1PhqdapGzrZbFeAGjtMK1Umhw0ShEHgs3xYuRqW61WkG8cO45XgaGniGQwGEQzUgQMgcVw9Hq93MK694PvgMEvlUpBKsKEV6vV8IyEBcTcbKEF2jJfsPheZcYcUoAjKdANSArvWiznRbnwhrwv5F7R+DojL2UHsWDkiMt90w7viDxgMF1xnespVj/ybqQbCfMwohgkb93FKVOMC2IQqM+4mCuMFFzR/v5+eHaUk7k7PDzUo0eP9JnPfEZJknWmwkiBOD3ThGwg4y4zrMNfSiPwna4iE1pMo3ipsCuLGwaEYn5+PjrQosxHR0fa29uL+BAvMhwOtbq6qs3NzShImZ6ejtZQW1tbarfbsUDOpPNc5yRcSYH5Ur53PjsRUUAUYTAYBDvuZOR4nO2UQ/nxWv6H55TL5VzWBeVGkD3GhwshtPCqNZTD068QZp7PdrTjHIdvkvEUp6MkjAOfMcfULIAO9/f3Q7G9doFyZuTFSUB+FwQEt0P7L1LBIAn+eB9GUBXFX75j0vc/bG9v66WXXgoD6VWRXkiFo3EnwXxhCD27830pFkouPnjkf5L0H0k6lXRP0s+ladpJJm3J35L0zvmvfzVN05+/9CjPL0cAnq8l1w4P4DlXSbl4a319XaVSKarHnLFtt9tRiEP8jGWmvfZ4PI6qtAcPHqjT6cSzvASXRULQiBtRBhQDgUUAUUi+46wwwgiMhKMgl46Akt+mCEXKthDj3ZgDZ+nxUPQwxOOxPwABpiVbMTyAHyE743tBpIx7kRQKDHJwGIzBQclYZ4qAPHXGM0AGIBgnjJvNpnZ3d3N1Fxhaz+FjoMnMLC8vR5twipkIT4qZjVarFVu+4ZEoX19aWtLjx481Ozurmzdvxn1ICxOqPH36NOc4HNUiY25Yr8oIPE9TkX+sZw8e+VeSPpOm6WclvSvJd/ncS9P0R87/XJkBkD78UBInT0AFngnwEs1arRZwMU3T2JhD3TZ9/lgEBAlDcOPGDZ2enurevXva2NiIGBfBccEGkXg8x3iBpAgKQo+3Bap6QRRxKSk5YtdarRYHbnBPyEiv78e48H/+8F2gp6Qg74DiCDuEJPPiFZJ4Td8UJWXenb99zZgTwgIPWyDbJOXui/KSkvSUn0NmPD4IgXFjkD1T43UMKB88CaQgVZOER17rcHZ2puXl5SinJjyDWGQXIdvBMZ4eFvluSV+fYggACXpVIcF3RALpBQePpGn6f9l/vyrpb17JaL7zWHIGAIV34k969swBPBoEYKfTCQJLmsR5T58+zR1H7jE63rfZbGowGKjdbgcM9Lwu0NZrAbgHlV50EELRi0aBxiB44Hq9njNG8/PzOUISlLO/vx+hA55ubm4uSpqB3hgKlNzbhjEOYDSKxHsyjxhbOBFJ8VzgO7v/EGoO62AsvBNhgLP63ocBb4+CMz7m2ecczwhqwbMT+pCCK5ezMx08LKLqk7CsUqlEmMEYfZ8Gz4FnoLEKexvIIlBT0el0tLGxoU9+8pO5DIHXa/DOjMPl2nXADeZlr6vgBP4zTc4k5Ho5SZJvSOpJ+jtpmn7lol9KnvPcgeLlxGDRu2Ll7RmSFEqxsrIS6S6sL9tVUQDqwFFequAol71//350zuEseU7wlbKuv0mSRIUf8BDFwQMQXngsyAXp58VAGAXfY+F73+m2AyFITIuweoYDpWZOiWelfM2G8xOgCYwkc0f6U1IutQk3wn3wxoQCPs/OWzDOmZmZCNtqtVo0ewF5oICgAzcYhEMURU1NTcWOTO6JIoPAPFQYjUZRQ0K2iN2afJ84nfp/jNjS0lIYau4HWtjf31ev14uaAVKnOABOTwKFgc7cyDkncBXXpYxAkiT/naShpH96/tGGpJfSNN1NkuRzkv55kiSfTtO0V/zdNE1/XdKvn9/nuQMblKYwjpyAu+f0uArPT4lwqVSKvPj09LQ6nU6k2RDSpaUlraysqNvt6oMPPsh1tyF+d+iOBwGys1iMDaGvVCrRNNO3rfrGFlKC/nsQgXgExo3yQaAxHyghBo5qNYTYU3uEGii+560lRcxPmAOULnp3QrFSqRSFQjD+1EBcFC4UaxeK6TI3bLwfnYAwfBgYlwdScqAB3z/h+xngNdikxFkRPPv09DQUnI1IzCHjZZ7q9Xo0NuU9SNM+ffo09jl4F6tKZdKbkB6IrvQu8y7XV3G9sBFIkuRnNSEMfyw9n4U0TU8knZz/++tJktyT9Iak/+fyQ51cTgZK+WIK/5wJBJIisKSCJKnVagWM85biKGm5XFa9XtdgMNDjx4+DlALGoox0qZWyDUR4cMbhYQwMPwQdYQ2Gx5Wc30+SJI4icwHCM7OjjXe3dZKUP6INZSOm9YsxwczjFVF+oDEZBucGHE05qvCiJs8ioIx+P97b+YMkScJ4sC5uiJMkCVTFGElNovh+XwqEqOFwEpOwgPSeQ/PRaHLycLPZzG1AclS1uLgYKVz6TPR6vfjs9PRU7XZbjUZDq6urmp6e1t7eXig6laqkpJ3P8VT4VRCCIQ8v8ktJkvyEpP9K0r+TpunAPl+RtJem6ShJklc0OZn4/SsZqV0XwST+uJfwmKnRaMQ+eQQPgTo9PY2aAY+nyQLcu3dPBwcH4RnxaggpngF0AUPvFhwBZvHm5uZypcMoE405gPhAa+Js3hPPR/Wce03ift9VCSfhNfbcF36C+eJvD7f4G/TBvGKUPKxhHhmnp8tATygnHtmNFGPESDnX46ERpCdzSOqU9adYyrf8lkqlOD+B1B1GmLEhU2R6HD2cnU0ajR4fH+vmzZuxuYh5h2D1rc8gJziRwWCgnZ2d6DQ9Ho/V6XSiQY0fiGO6lUNJ7lQuez1PivCig0e+KGlG0r86FxBSgX9V0t9NkuRM0ljSz6dpunclI7WrmBlgYjyvitBjLBqNRhTBIEjz8/NRZkucKCm3L4DjqCQFTOR7pMM4psqVU1IO4mJcYPjpb+85d7xzsXYc5ZubmwtG3AUPJEQsTGtr4nIUGg/oGQeeyfsQCtAEA+/paT7mxu8lKRcmFDkH5z0wQl5K7ZkVFNq7IqGgrL+nUXmGE8TMJwbSeQf4Bp6N12UOPDXMfT1kkBSHiLz++uu5EvRiCJOmaRQDSVlV59bWllZWVnTz5k1NT0/HvhTqVpBn5s3Dge87EkgvPnjkNz7ku78j6XcuO6jnudyjnD8753kRaAQPoo2fVSoVNRoN3b9/Pwig8XgcaaA0nXTy2dvbC0KLbaTuaSF9nLXGEzFGZ8ElBWGEYHncz7N8QxP59cXFxdj5SCUjYQfPA6ayxZn7MhaKXBByDJZDZUkRNmF0QDjk4vHweEcXTIwfCoNH9oyJn+vgRpC1RVk87YtSeFci+BiMCyEBSMNrCUByw+EwUsGVSkU7OzuBBKnl4N5Aem/Gwnzt7+/r0aNH+sQnPhGdjD1r4DJAz4Gzs7N41tbWllqtVjgQitdAlo78XLavigvgujYVg8XrIi4AKOqFMQiw58uTJNHa2lrUgXNSDA0gEQz2EeBRJD3D+DvDLGUFPgiCN+uQFLDQjYIrPiSRe1+EAc8JHId4ajabkUp0/sFTSSi7GwMvxMEo4LVgsxFYTwEy3yhqEbYisDzXi554N5TR15L+CCgsRsNrLbxuA0Vl5x3ZCGSgXC4H2vO6D+aTMWFY2J5N8ZU7BNaF++FgDg8P9cEHH+jTn/60Go1GOA0/hAXjCumHQX38+LFu3LgRmSa2gpMl6Ha7z6Bd5usqr2trBNx7ePUUVW/OCxDz4hXSNM0dM45nW19f13g8OaiTuM+PsaalWLFSLU3TSP05k+3FMAgcBSiMwwlJZ7dBGZCXZAUQrl6vp+3t7djWjILxu747jvEwbikTJEcAhCvU/FMoBKnHeD3E4f1BO8w3dQ4YLt/DAfQHdTAOL8DxDAkhHO/qFYVeBwICkhQFRCAFjAe1GtQxJEkS7cYpsgLVefjiexBAB7xrr9fT/fv39fLLL0cFILwQoQ5EZrVajR4CW1tbunv3rj73uc9pbm4uek5ANDsxeBFZfFXXtTUCUr6M0uPO8Thr/MHCSoqF5xTcNE2DYV9eXtZwOAzFRzBQYI4Sc0IGK895AyiZl8s6vCTFx3ZWhMWRAcUlEEzurVBSnktlWq/XC8SCkk5PTwfU9HQSyu87Dvmb3/f0ppS1V/eUGuPFmBVRF6ERtRRe1cf8uZIB4UEYQHP6MnDx7igiBpj3ZkyeIsbY+CYyxnt2dqalpaVoGELBFWvmaM/RDQiLd93f39fh4aE++9nPxsag8Xisvb29yCBQek4D1JOTyfHlw+FQjUYjuiLxTo7siqEv834V17U1Ai7YDn9dcPmO7y8vl8vRIZhtv8BjOsc+fvw4EEWz2dTy8nJOUEEfsPseV2OIsP7kgfkMiI03lRRbkb16Dy9G7trRBmNGaRB6hJW8/NLSUhRC8X2E1+fEQxtSl8USXEmRwvJ55z0Yn3tlvKXvjyjyNsTRwHT+YJDdsLsBPDs7izQpBmZpaSnQAPcol8thiJhr3oV1AY7TR6HYh5GMA6SrbwLD4A4GA+3v7+u9997Tpz71qdi4xFriACRFRmpvby9OluLkaa9C5FlOhDN21uYqrmtlBIpxvsNtZ4mljAwDiqIcEHrkiGkHhtA8efIk4Nra2loQSM4UO6RG6L2LELGdNDFGCKqUnfnHu3iFoeeKPUfP2GkRDutPmrEYL/ouOroCVSqVKBJiziAc8TqgBpR7PB4HdEbBMSLsxGROHD2wi9C5Dt+oAwPvewOYPxqw4uUpz3YI7CSsF+xwaAw/H41GWl5ejvlivMXsCvtHXNFRXmohfM08Bem7KJMkiR6Db7zxRjwXw03hEQb/9PQ0SolpXTcYDHR0dKSlpaXgBZALz7BcRbkw17UyAm4RPa8sZdmAiwoqMAbD4TAOsEB4KpVK7OUmfyspd2oxzxuNRsG6IxS+F/7g4ECDwSC8mcNfxkAKiJgbQ0DvPoSbz1FwNioxXk85shuRCkRiYDwpDVBAL572o8MPQugbacbjcYRAZBV8K/Hi4mIYUYfIGCYIVK+Yo+eBKxHFWr73wAuH8Nje6ARUQXiCIWb+mW/vv+DZDDgJwj3WDxQEB8F1fHwcc0G4QlrQaxWSJFGv1wC13JUAACAASURBVNPOzo5u374dhhhDMBwO46yHpaUlnZ2dqdPphPHf39+PRjeURHNvh//IzlVc18oIcLkRABU4i+qeEagOqUSPedIxbCg6OTnRBx98ELByfX09DIXHipQNcyy1NIGYW1tbsSfB8+YQbQiBb0rysMILhBBQdqIxTjwXrLG/O4JyeHgY3pOaBDgS4C2kJQqLEqFgGCBIOCcVSceh3NRfYBClfHWihxYYK1dQPDalv8wLY2XtQAMYWFqteU4fhOJHpHmbce5HL0WH3/Pz81pdXdVgMAi0tbi4GEaE7ABHim1tbT2TIsVBnJ1Nmo1yijUc0NTUVDR9Ze52d3fDKCwvL+vx48e57IYTsB4C/sATg07u4LHwfsBrZ1P5f61WCyIHiF6r1cJ79Pt9tVottVqtCA8g9Hq9no6OjtRqtaLQh3Bia2srDrPEW1IpBoFDig9uAITgpcWuXAj2zMxMbHFm3Hiher0exUm9Xi/Seb7PwO8JOiBPjxINh8M4Ppt5Q+lAPb6ngTMNqtWqpMzY9Xq9eEdibtbJC7JALxhGDAOIgUo97j0cDiO+BkbD59BABaWgiIlDXRyBwZXs7e2FIQBR4CQWFhZiLpgz9lNwH8KmjY2N3F4OxlIuT/oevP3223Eq0dOnT9Xv96NLFbxUvV6PPoweusANufwiRy7bV3FdayPgXteLSjwkwNoPBgOtra2FwUA5EaB2u63l5WXV6/XwTFhg7xbjhTCPHj0KFMG4iJ29+g6SEYQCc4ziextuiCu8JvUMeAJY6v39fY1GkzMJHGryczwXAsV7IkQorqRQMNJxXgAFaQmaSZJnTxjCW6GgoA2+j0LDEUB2Yfzw2F4piDdnPkEBvsZS1vegCJWB0rOzs7GBqFqtRldm2nxhWLjv4uKiFhYWgjDl/ZEFEMf09LQajYYWFha0tbUV98Kg4PVJHZKJInPU6/VUr9cjq7S3t6darRa7VZGHSiU74k76zgfyvsh1bY2AlGfj/f94cKx8mqax6YatqAixx5n1el2SInfvELbRaER77Z2dHW1vb8e2VATIiTWHqK4MKKmTS4xHUoQF1KxTPcZz9vb21O124xkoqBOfcAM838uDeU6plB2A4jsBCSWoZaBDLgaL3+eQTd6F+WQeUHDQkJRVUuI1vS8A84TRckad+nwPgTAYpVIpjCzKgiFjbjEiXkMAgiJliwcHHXjVIfKFLPjPFxYW9Morr+jJkyc5shLl3djYUKk06VnpGRx+Ts+Bg4ODaINPCANiA9Ug+/7nKozBtTUCKIGUCREWEi/h3pv93Z1OR4eHh9FHECFzj0p6EaFl+2ij0dCTJ090//792CLq38FrUWKMUnK54CJ0vhkJo0M8Xq/Xc4vPd+bm5tRsNjUeZ51q4Q74/SJv4rULjp6kbAefNMnre9qUMIvYmL/JQjifAaTmPT3sIVWLQeAzSUFgesqTsTI+1pfe/9yj2A+CcXvxFEaQsVMpyjuPRqM4GYr7eqkzaUAyPPyNQs/OzupjH/uYHjx4kOvGlCSJtre31e129cM//MO6ceNGjvvBoFJHwJp4JspJQXd2V3ldayPAZBVjJq8Rl7IuPngNKeveAnRdXV3V/fv3g9XudDpR3LG8vKyZmRm999572tzcDPINmF2v13MxNsLuqUQINToUeSzKcWGQYzMzM3F+IEdok1pqNBpRv09qr9VqqVqtRtqNI7K9AQYGBM9IlSR5fxQXnkKahAnsm5+bm8sZBUc3lFp7+sy3ZRMaoBxSpthwDsT9KCpkGsbbPTDrh3HzWglCizRNIz0KF0FLNsqj4YIYJ2EIaLHRaASCgAPyzAROZH9/X/V6XZ/4xCf06NEjtdvt4DW4F/sE7ty5ExwS5dkUsMG5eI9CP/SliICv6rp2RsA9BBdeyAUQ6MpklkqlODIaTzw9PR1kDfvAKQQhB1+tVjU9Pa179+7pwYMHIXTeMATrTROKIhT3zSqU52L9aWpKGTMeqdlsan5+PurN6ZjraMd3NeIdMBpSxgcwBmC4ex32TExNTQUhxnx4+a7Hung5xi4p+AEMzHA4aUjK7jjSpZ5ag0sg/gd+E5KQHcGge1qR+4E+hsNhLvVIpqDT6YSx8EahZHuKlYzUBGxvb2tjY0MvvfRSdDgCOfm8SJmRJ87v9XrqdDrqdrth4J48eRKOgnf01KijHg9jkDFkiTE4orvsdS2NgENkILnHvUwYCkOjBhaySDgdHR2p0+lofX1d7733nkqlUmzimJmZ0ZtvvqlutxtxJBCfcUD4oIxYcPYaeNoSQYcb2N7eDuRCSEJ6jwqzfr+vXq8X9/C8O//2qkVXEshJJx69UIZqNwqJgNMUI7mAonCESxheoC25dr4PAcj94WRo6lKskcCj+zvgFcfjcfQFZG4lRbrXUSFkLvfgIBGMs+/yYy8IY2EvPxkLNpGtrq6GAUMR4WPovEydAqk/9i2ATvr9vhqNRhSngUIgW5kHKas9YJ29cvKqycFrawSAQ0UvwSJ7rTVsL1CdRZidnY3cPuTYa6+9pk6nE+mab37zm9rc3MzxEAgi94fxHg6H0VmIAhpSa76hB+/oRSZ4KYwG+9U5zdh7DmIoKGDZ398PVIGwo/x+5JgbKSkrP8V7g36YV55H1R9xPdC/GPejeDwP4SYs6nQ66vf7ubp3vBveznkNDN9Fz6PzMWGJ907wYjD3qp4NIfUqZTG5/878/HxUjrL2N2/ejJ6CGITZ2Vmtra2p0+moXJ40MB0MBrp9+3bMnVejzs3NRXhJkxEMNqFAsc7FeQLk/qpQgPQcLceTJPlHSZK0kyT5M/vsf0iS5EmSJP/6/M9fs599MUmSu0mSvJMkyY9f2UjzYwpLyASyyAi7N9MgRqYLD3AbIfPiDEi3arWqe/fuqd1ux6L4qUNAazbcNJtNra+vB5sOGnEYOjc3F4xvkiSRVkqSJMKO0Sjfc1BS/MyN2fHxcSiYt7UmHuYPRgYjSVxNyOQhTb/f197eXmy0IvbHa6JwxN78n/CCNlpeSUc9Brl9DIA/mz0YwHRJcYbf6uqq6vV6dIaiWQoK4cqA0u3t7Wlqair2WgC5kQdQBdCcHZuEatR5eIFSv99Xp9PR2tpaFI1Vq9VYj+FwqE6nE6Tl2dmZbt26FcYe5ESDGj+9ilJhkCZZJN8C7bLP38Ww+EWv50EC/1jSr0j6J4XP/5c0Tf9n/yBJkk9J+k8kfVrSTUl/kCTJG2maXl2hs/LbYD0liFGACJQyAmpvby+ElQNGqFN36Lu8vKy5uTl99atf1TvvvBMeGEIJgomTh+v1umq1WlhrhBsmGwWn4IYY2DsQ8w5JkgSnwBh9ExEn31CFh9GDYwB1OET3P1Lm/X3OKCnm2HY88Wg0itw698MjS4rwwctsiVWZd0/14WX5XU/1wTV4TYUjBhAchCfcBxkfCLh+vx/lt6RYOabcQydKmBknVYHuleFUmPejoyO99dZbqlarWl5ezh1uCrFZ5CxWV1djp+d4PNbW1pYajUYcdcY+C2QDWfE0sxPMzOn3NRxILzh34Ntcf13S/5FOGo7eT5LkrqR/U9Ifv/AIC1eRE7joQlmJ8cbjcWzEoBwU6443npqaCmW4f/9+pL9gw92D4dnxCnga9hNIym2gQdnOzs7U6/Vy95MUCi4pYlSUnN/3Elq8HFVvpCMRDt8jQP0DCuyknpSdhOSHapD9aDQaOYPV7/ejbbe31ObntAX33YJ4StbLIbD37UP5+V3CE5AFn4NESAP6Fu5Go5HrJkTsDzpj34JvF8YxSBNDz/F0GBpCh9FoFKnYqakp3blzR9VqVUdHR2q327nTi6jBQFHX1tbC2SRJEqc2Y3z85GPWhOcyV84reUbsKq7LcAL/RZIkP61JJ+EvpGm6L+mWJoeRcD0+/+yZK3nBcwfOfzdHAiFk/IyrUqmoVqtpeXlZDx48CMFbXl6O/eMsTLVaVa1W06NHj/Tee+/p+Pg4rD27z1jU5eVlrays5CAcGYgiV0Gqqt/vx7PG47HW19dDkDESo9Gkgw/lviggHqHT6YQws/ttOBzqxo0bIVyMg2wHBgOuIk3T2PsOdCV1B1tOWq3f74dBwcOxocrRCNwEqIDvLiws6PT0NMIw5uaijV7upTGqKALGgbSmIzLCK5AaionBJtxy4zgajWKuWFPe0bMXfu4Bf6gfeO+997S2tqbV1dVI72G8QBpeZXnnzp04z7LdbscuQf7msJLFxcU4W8LDOvZZXGVqkOtFjcCvSfofJaXnf/99TQ4hee4rfcFzB1hQ96TE314OS6zKnnMU8MaNG5qbm9PTp09DiTlcZGtrS1tbW+HdECopKxBZWVkJ78mzvIMOQkoXXrw4THqlUolNN+SRgbOM2zvhIMjSJOYmbTY9Pa3d3d1Qci/IgYWHIDs4OFC1Wo3fQxlIw+Fx8EoYPU97Ei97fYR7o3a7HR6a9fANQTyXP06ssq5wORCfaZpG5RyobH5+Pojds7MzbWxsxHP7/X6OqIQfIVzje7wzhKojLklhvCFX+/1+OA9pYhy63a7effddJUmiVqsVxnBhYSFaiBGSSBOU0Wq1tLGxEcYMnmdqair2M2DEQQPIAo6iWCDnDvBFrxcyAmmabvHvJEn+V0m/f/7fJ5Lu2Fdvn392ZVcxDsJSAsMQXmAUsSZwG0Gj9JZ88tnZpDvM5uZmCCy95mq1mg4ODiJ1x2LA5uJRIbpcuPGyHmezdRSUgLJxTiL7yQ8PD2ODE/sfqLGH3CqXy8FMc2FknG0mJejj2d3djU7JjJXn4u0Zk/MWkqIYhvoGBJKMBt2YvCko3pswwivnfNsu/8dQoMAgFdAT6wonQkxPag5ik0NCMeTSxOjQ4ptwRFLwCOwJAXlRyAWMB638yZ/8iT75yU/q9u3bWltbU7fbjZSgFy+xlsfHx9rc3NSjR4/UarX00ksvRYgGamAL9u7ubq5KkHnwd7iK60XPHbiRpunG+X9/UhKZg9+V9L8nSfIPNCEGX5f0J5ceZf7ZuTjJlR0P5D/DCEiKOBjvNBpl+/ufPHmijY2NKATiqG9JWl1dDWXwrAQcA/8GDkO64a3wbjDDQFL+JofM8djEh1QSojCuhCg3ZByKRogA98DpwjTL8LDGS215hjQxepzQjHKCrBgfRhDjwH4FtlhzohHcCKXbg8EgttOC3lg74nBCEknR4o3xVavVMPTF+hCMC12hcQaScjl+UJ5fGEmIUgwfaUIyCZ1OJ8IQsh5vvfWWDg8P9bnPfS6MPRuYyELAtVBARXk6xURSdlALXIRnCEDAfyF1AsnF5w78u0mS/Igm4cAHkv7z84n8VpIkvy3pzzU5nuwXrjozcP6cXJxG+gflwSt5hVWSJFpZWYn0Dh5kfX1d3W5X+/v7Ebv2+/3wYiwSDC57v/GyPNPz2IwP2EceHqEtHuuNUJbL5VyTjsFgED8jtww8lrLUIZkOj9/x7qQTmQMguntXJ8c8nYgBBTJjLHiOM/u0J/cQCWVC+SBUHTUwZxTbkNKDx2i1WhHKFeQypyTeB4EmnSAv8vQgNSdrQQ9wD5Ji7ufn5yNel7I9Dqy/G8IHDx7o5OREP/qjP6qZmRk9efIkV1OBU5iZmVGr1VKn0wnjgNLTb9JDL995iOJ7hehVXFd67sD59/+epL93mUF9uwthYCIuiitRSOLefr+v9fX1XA092YCjoyM9efIkGj5yUb8PaQRDXa/Xo9oLgSCGhZzz03C90gsBx8uQo56bm1Or1QqmHCHjHSqVyRl1o9EoSk+p3vMSWJRTUvwfmAtZxd8zMzNRxUf6kJOEQBRkSOBFPBRwg9PtdlWtVmOvPV58dnZWvV4v57Xw/J5SBVkQHvV6vTCSaZrmyqMh4RhXuVwOCA7PQjzP3n32P5ydnQVPw/0xJsV6CDr+siZO1JXL5SjpJrQbDienGd27d083btzQ2tqaNjc3I2zzCstyuaw33ngjyNe1tbWQO0Ip5sjn7Xt1XauKwWJ60BlmDwFgj71QwzvtkI9eWFjQ06dP1W63c9CRTUGeasTbkZrb3t4OFICHpmc9Sp8k+TPyUBw2HBE+oDhYewpUEAB2DFJwdFFWhKYjnlaSslOLHbE4f+BlzXh6NsnAtUxNTWlpaSkM4dzcnE5PT2M+KK7iXTBwXnLMs+AMIFB5V+6HNyT294o56iikSdUfYQV7HkqlyWYqGn1QBOQICO6I6kRCJd/DQCHP4uKiqtVqlIxLCsSyt7cXaBGDcnx8rG9961s6OzvT7du3tbq6qq2trVgDCMharaa1tTUNBgN1Oh2trq4GCQqym52djbXhuVK+VuCqrmtlBKR8LwH+xru7cjijDKx35n52dlabm5uxtVjKmoDi0Sjd5Z5sqjk5mRxTLU28/ebmZo7gk7JiGLoCET64pyR+9D72/C4IgypGIDNkFyjBnwW8RIidnecCNQBP0zRrb13cx05vPAg4DBlQnRh/eXk5+AsvTKpUKrndjexTwCMSapBVgUPhOXi/TqcTHtybkzJHvo2aMWPk4QkgdH0fA5WTOAvvR4AhJJ28vb39TLs5GpSyjoPBQAcHB/rqV7+qT3ziE3rttddi3wIyhBM5PDyMdDDjJ0xlvG6sr1rx/bqWRuCiyXDiBIXC0sP+EzsijBwjBbSniostop7X5hkILbC83W5HTTzfIQyhTpxehCwkvQRgn/k+xBvklBON/M3v4xlAHLDK/nmpVAqv6xtUvGcieygcreCdQUCw5/yN0lar1VzbMw9/WA88OeENhKDvomNdGDcpSFhzirDgVTB2TtRCpGGMQA3dbjfIU89MkL7z/QakTAkL3PvS2IRKQ4qUmHfembbl9+/fV5IkunXrlkajkZ48eRIhphO8nHlQPMYMbsArGZ0Q/L4Sgx/Fy9lwDweAvVh9Z78p6nHIev/+fe3v70dRkbPhsNX8W8oOr0DwHz58GIQcMd/S0pJWV1fVaDSiWMfjc1JYKLi/A2y6lB0RzsWCM3ZQjhsdSSHI8BvEl87ugyjYz87cUV9ADM7P8GZUFvr2XuaLe0N+YSx5P9JsGFaYde7B3GIEvPmGH8QiKYhbkIGHOZQ4Y2h5djG0Ia8Pv+HGFIOIB/ft34SbSZIE9yMpUABhaZqmev/99zU/P6+1tbVAqlR5eo0F5B/pVOaCw27pQegFct9XYvCjdrFIUhYKODrwfQUI9uLiotrtdvwuqZ9Hjx4FAkDgWWRy5RSXeEVbp9PR7u5ukE1pmkbl18rKSmz08Tp5EApGgM/Ib/s2XuJOPiMGB4rzDhgqjBUpR4QIVh3h4me+UQeBl7KCKEprUWznWzwXD4fA73vlHkbADR6fgxpQXtbMPbunDh1u4x0ZL89mDplXZIFtx845HB8fR4GY1wcQXuDxUfalpaV4Z+d8IDKlzEHQAIX5v3v3riqViu7cuZOr7WDM1CV4mzj6H4KCXOG/F4bgWhmBD4uJmJiL0iheW+45fQ4hRUBRBhYRko0CGmLj7e1t3b9/PxBGpVJRs9lUq9UKNMFC+u5Gxidlpw+7V+ZeKB/5fXiM6enpUABCCT5n3OS6MSDEvHhE8u8oNsrEffh97o1RZN54F0nxu8Ph8Jkj3fy+Hp55w04Mi5Tnd/gdR3elUilOBSac85CC+gTvF9Hr9cKTkt1xlMF78PtwKYQ/GE9Sd3A3jAe4TjbIwxQMNRvXtra2dOfOnVxK2TslwavA7VApyCYyJ4wvQr+Xva6VEfBYiEkopgyljEnlwqpLWReYjY2NILf6/X4sMPEpLDGkXbfb1dOnT/X06dNQpuXlZTUajdjmCpz33D/oAe/kOXYYeeC9x9OeIiSNhpGA72BOQC8IjvfLJwzw3D+e1Y0mc8h8gnwqlUpUSzqi8RDJMzA0bwH6gwC4F78jZcdxEWqAHuANWF8MB3sCOBMCg+6pVU8/emUoCgQ6wShRjAQMRwZYH69C5R2ctMV78/uQ0t6h+uTkRE+fPtWrr76qRqOhdrudK20vziHt42lK41WCXD+w2QEX2KJVlPQMGkAo+/1+LF6lMjm8YzictL8CzlOY4wKOAGxtbenRo0eRv240GlpbW9P6+nosNm248US+cUXKmlLCEXiumuIWnk06U8oOUUUAEQjy2MUjz0AWeBBiW4QTTwhRRhEPsb/vmmQ+fa8El4crnK2IYFPH4BWF/Jz3dCIQ5eHfbvRQYA+BPI5nrYHjGBCMEGGFP0vKGokQ13unZH9/whHIYwhbjBcZhMPDw1z5No1l4Ek2NzdVr9d1586dXBoUVHB4eBiNSOhzSOk0iMNrJ35gOYGLMgNu5REMFn1hYUHj8TjOr4dM29jYUKPR0ObmZm4jEopaLk8OsaxUKtrf39e7774bC7K6uqqbN2+q1WoFYQXKwAthPDBQjMm9OArD76OEUmZ8IPxQHt4ThapUKrkGG3gsnokSnZ2dRW9F5pFx8se9kdfhj0aTnY0QoGxsoc+AIwp/Z3LzblQ9npWUK6n26kHmoIgKMJb8DNiOgUKxeZZnKdzjosQokxOWELLORxCjS4rQQlKEDzwHHgnjWKvVwpiMRiNtb2+rWq1Gh2gPCzDSZEUIdeC0dnZ2csr/A88JeK1AMVbCe1FN58Uvfort0tJSdA3yTMLh4WHsS9/Z2dEHH3wQ3AKbRKiOw+jAHeCtXQBRVod0XkrKeL0eAMGUJpuBuB8Ch8ByedkxW3y9qzJjRbEhFFFaPCfklKTcuDxF5YVEDvUxiF4V58gBNOFKgVLi/YoKLGUG0bsrE45xJiNzyfZo0JCjPH8nakX4N15eUhhLOBjGzLjhaLhYXzgE1oN/O1ewt7enpaUlffzjH48mrWx28q7FhAQYAe89iNyjB/8/J1DgBJxQIifsjDP5f2A5igV8p7CI4pCnT5+qVCqp0WhoZWVFrVYr2lsRBnAvZ+pZGAyRNxjxKjqEygUG+IjguHGTsr70fIaCcP8ihEZQqEPwralAaK+0pFLP01ij0Sh3cIenPimoKlZrgjbw5nhVf39QghdBOZrwz0BM/GGOSQtKeqb5CO8CYmA8rA2hGL8LquCZZBRAh9TysxY8i9CMNeVnhDS+qazb7arb7cZJRxCPGIHV1VUtLy9H9slT1e4AfmA5AenbWz9XAspYvcEECk9bKBYWjwYcf/z4sTY3N9VqtXT79u04oJQUnsN23ypLTp5nUSVInOhxJu8Ak4znYpExLr51GIOAwUKZeAfYe/LikkIA+b/Hx8wlYyX2BeYyX165hjK59yMk8Hy7x+soJOPg3swXCMKzBUBlagPI0bO5iHoA7u+bjBgvc8/efUdI3mW42P/QDQdNRSFG6Ujl4Rz9CVBWHI9zORjgbrcbTUWoygSldbtdlUqlOHSGMSDTjpBYi6u4rqURKP5dnIxSqRQlvkBMPDglwkdHR2o2m7mJlqR79+5pd3c36rvpB7ewsJCD73g/FM9hvxsBlNM3thD3YQxQKLycC5gLOZDbWXAp2wbrXIKjIuA+7+j3x7BxHzwuOWx/P2JpxiopNiIVGX3PNDhk98wAKIGybu7hGRoPgZhfz66g4IyfcSIbXt+BESgSyHRGoiIShaVUvFqtRg0Ez9vb2wvZIYvjKIb39orE4XDSLandbqter6vZbIYssAlLUhyNRyk1IdtVpgX9unZGAGFyxXeSBM9FrIWXZJ96p9MJUqhWq0UKaWpqSu+++676/b7q9bpu3rwZZxNiiVE80kMsOmnEYm4b7+UWnO/hPQkH3OPyLL8f//eWYO6VEXby14wRpeYdMEw8R1LOCBISoBxAXja3wFp7xaK3NQc94N0xFu6lPTTivalKZGww6yAj0o4oI0QgtQfOWTjD73POvzHIvq4YHEcL3v0HYnJxcTGOpXMugb+ZcylLB9MObmpqsllpb29Pjx8/js1CGxsb6vV6uWPlKVLzSlefxx9YYlDKdxZyAXejALNPhRj99nZ3d3NFQpSQHh0d6d1339Xp6Wkc6VWtVmPbKkKNYZEUxBELhLUm9vX/O1xnzHg3jIlXrGH9MTjEzF4shCICqfFcXlaMcPMcvJ3DcgwARVXFY9rYBuvpPh8/iAaoTlt1RzBUMuIReR+UlfWglp/nYRxnZ2ejiSnhl6SYJ1AaeyIwwlJm4JwXYh2YA9bIwy5+9+TkJBwIXEG5XNbLL78cWRfamjEuUolOdGIwed/d3d1AmawDpcfr6+uqVqs6ODiQpBwJ/WG6cJnreZqK/CNJ/6Gkdpqmnzn/7J9J+sT5V+qSOmma/kgy6Ur8lqR3zn/21TRNf/5KRlq4WDyHn5Iib8smIcifmZmZ6O+Ox0nTVM1mU1/72tc0Go108+ZNSYoSYPeaMMqk7CqVSq7AyONiz1b4GD2cwPPwuwimlMXEGByH0hgA9zh8h+dLWfk0CgdLTWgiZeQYCoggp2kapcMeIqAAvAuKg/JRhIQyc7E2GAFCAeZRUoQCnmYrpkQZP1wBfAcKBl9S5CSKjoI5ggwEsfFOzKVvLOK+3KtSqWhlZUXSpL8iJCJNYpk/J20ZJyEA/Qa8QQvfR7YgrUGAjO0qrxc6dyBN0/+YfydJ8vclde3799I0/ZGrGqA9J/e3K5lnCzxmp/Hj+vq6kiSJ1tyeFmQDEfsHqClHyPDSUnYGHd7EvYZ7eicIHb65B2LxMQh4aITaCTZHGn4gBf93L0EVG8/1PRG+qeciHgUvS/7fMxMYtyJR6GQekBf+gawC70HoQjdj0pne7QluA0XF2MGuY9RBC/wMzsCZfc/U8C7Ou+Bh4Qy8tJiCHYwTF+/AfZ3Eoz/AaDTSw4cPI6VJlgBEKE0Q1oMHD6I5CvtDfOMXxsI3lhUzY1eBBi517kAykaSfkvTvXXok3+XlufjzseS8gTd3bDab6nQ62tqa9EdF4KkZuHXrlra3t+PUWsIEmGgUkngUT4Vg4anwpr4HwIkh4lQ8LpV2KLZnAzBAhAMIOCjANH7g0gAAIABJREFUe/h5KIAhqdVqMS+SchDeYXEx6+Df8T0FUoYcfIcjQu69EuAJOCDUeRxn9BmX11DwPYylP5/PCPcwll63wO+SFTo7O4v58XSb743A0DjkZx49rKJSkbXGuN24cUP9fj+6P52enmp1dVXb29vxzCSZnMuwvb2tSqUSHZf29/eDeGQMzGe9Xs/tVCzK+1Vdl+UE/m1JW2mavmefvZwkyTck9ST9nTRNv3LJZ0h6FgZdNCEIrW9lXVhYkDRpTIHHwDvRx39ubk63b98OpfQNNig/R4X7dk+UECH3NKSUkXdeM48woWi+oQZPUfT+wF5gtMfBGBVO4HGPjaFiDKS7PHXlpbyEJRgx6uppIkqqjJ10FLxAnvE8CrO8Qo+UnmdKHGnwmTP7GEBX9FKppF6vp9XV1SiSgsBjvp0cZM49pJEyBMOzGLf3FMD7evGQE4gYnMFgENwSHYcJIR4+fBghAKHFxsZG7D6kNgAHAyKcnp6O5iPIA+lO5P+jQgz+LUm/Zf/fkPRSmqa7SZJ8TtI/T5Lk02ma9oq/mLzg4SN4qiLURsDYrw1MZgegd7SZnZ3V3t5ejpDDWABpiTtZDNp6Ua9PDAxj7fGkQ3CH9nwfOOrQFKGDkPOimmLmA6Vlzz1xON5tNBqFB/G6BI+ZGQ8sPPOJ8ZAUKTDuQZgE+QmiANKiHA7T+TeZBa+RcPTm78c7EpJgoDymh38grCL0wPBgWDFovA+oCuPh81/8HVAga808SYoQkbSfNOGScBIYorOzSWt35On4+DjWplSaVDnu7e3p9u3bgQaczPUKVM8yMW9evfii1wsbgSRJypL+hqTP8Vk6OX7s5PzfX0+S5J6kNzQ5pSh3pS9w+AhCw7+5sOIoMwLearWiqxCHZWIgOOkXoZ6enlar1Yp4eGpqKrczUFLEsgiVd/F1zy0p58WcGHMGGa+Hh0VpvPkIBUS0R0uSbM+6lMXMCDb3Bm2ATrx+gXQXO9SciadDMGlViDopS3lhUCDrEFD/HiXaKKkX4zipScjE/72YibXlfjDroCXPQDgJC1piXjD+PJ/4GyISZ4ARRNGRDSoRPUxjPhkHZxv44bfj8Virq6uan5/X+++/H3LH3gzCv263G8iGxjbMP2Rj0fN/VJDAvy/p7TRNH/NBkiQrkvbSNB0lSfKKJucOvH/JMeYuLKCTI1I2KR5D0t8OxpZ4lVJQNwKSIjdL9RehBay7F7t4es0Zbu6FwoJSfJtpkdR0KA609XQanreYJuL94QMQYuA784L3g5Ng3BSnIOhedcccYwDL5XKkworFOE78gZwkRZm2I4GLmHvGenZ2FjDZEQ8eEsPsawzScphMmILSo2zMsyuxzxNrR+zP+mEYWV/2J7ihGg4np0IPBgOtrq6G8eaezWZT77zzThimZrOpnZ0dlUqlOCcTxEm/SmQCZMCcX/X1QucOpGn6G5qcPvxbha//VUl/N0mSM0ljST+fpuneVQ6YheCPlM/DEl/CIBPXooCUZ6IUxGt4xJWVldyOOu9kw+UwkjEB9RE6XyyUzwkuJ+RQJveGkFPk3YutyRAOeBA8le+kK95LUhiks7OzgLFcnib0WgMYauA8hUPMO/PHd+nBR+XbaDSKjsCSYh8/xo1uPxii0WiU25pNOMfz+Dlj9nFClKJsnjVgLNyT3ousKT8nc4ERoqLR0Yc7ADdqZAnYLk6x0MzMjFZXV7WxsRFGhN8FDYBoKRqCR2C87vi+r8RgevG5A0rT9Gcv+Ox3JP3O5Yf1/BdGASUGojWbzWDRfcI6nU4sLkrj5bR00B0Oh1Es5OQV8SgL7DXlvj/dEYrH4Si8w1P3KMSBvumFzjOcqsvYCSFGo1GcdEOcjsL6dmeP1zEMCCLGAQNGhgV+Y2pqsjUWT+skqROgXnCEUnBvlB4jyxgwAHyHcVKJ6CXP/X4/yDu4GwqJmFM3ZCinIwspqzNhnKAYUJHzRWR6+Ix5xXB4BSOHikAm7+zsqFKpRJ/BNE21t7enqampaGdeqVSiS7LzKdPT09HzwmXYuZRiAdGLXNeuYlDSMwvstQH8GyYX2N9oNPTo0SMdHh5GOywEi3+z6YgYVsria7wEzDcwGaFDKBEWxuDpL+Alz0QweRcq5/A2XiDjJw6jbLyzpFw+3uNaEBKKASR2ZeGZTsTxblLWrJTGpBgbjIOjDL8/7wzJhaB7KqxYDSllxhAhR+DdcHqGgZ2M/Mw9PgqDofP6A4/veW6lMjnoxUueGadvBvNQgrAR+SAEwoA6F1Gv11UqleJ4M044YhzO6xwfH2tpaemZQrIisXpZbuBaGgEp87ROeiFItVot1zG3Vqtpd3c3+s6dnZ1Fa3E8Gg0rxuNJySepMCnfK288Hueqt7wtlqSc4AJB+Rsv4u9A2md6elrdbjfXG4A0lDfPqFarQWp6ug3hxJvjjbyEmXF5hWKRROM9QRIYxKKRIUSBi5AUuXlgcpqmQWB69gFBZr7H43Fkc1B+MjMoNZt7UFiMrXMCkMCkc3k+8+7kLUYMmaGQzGv1i52IpawZK1kh6iEIR1D8fr+vjY0N3bhxQ41GI4w4BWyc8SBljUo4UAUk4NvMi5miq7yulREokoLFicFTsqHH+/r3ej31er2AV0BWDrWg1huBc7KOxpTAQd9+jFC7V8M6O7HmkBtBJIWIMgKzqQuQlCu4IbNBlsJ37+EJaRuOJ3HYyPjoRYhSELLAL6Bo09PTYQiBvayDe0vQFobF42+2/eL5kyTJhQu8K8aZeePCABOKuRfEyJfL5RwZSL0HBp0xsa7AeGSJHhOlUim2GCfJpOCJ1LLzSswb68PFOlCPMR5PWtMtLy/H8XUrKyuxJ4DuQo1GIwyEZz+44Fx4b66ryhCUvvNXPjoXL+3MOsIPnIJMogccp8BubW3lGH26AxF3nZ6e5spRp6entbS0lDuxpl6v5xQGr4g3wgv6OBwWg1gghiDCnMjjc97LoaqU1THgJUEyp6ensdEFQUdxpKysGS/urD739CKn4uYdlKTRaKhUKgVR55wBzyxmRLw4ibEwx74hCQThc4fSeSzvqVsPAUmrecEXxk9ScBjMH7Uf/rypqUmzFOpNqOYDpXAdHh5GabET1TwPw4GRcRniSPQbN27E+66tranX6z0TckHQFrmM7ysx+FG6HEZKWUzk8JZ4/eDgIHaw9fv9qA5EAdI0jUMlETaIMja/kJ/Gc/lpOXgij6URvGJBCYJCGgjvjlJQgIQhYdFhzv2oq+Pj44C9eCw8PltWHf4zXx47ktIiXIAA9BDk8PBQS0tL4cGBxd1uN+5BOERtQ5pOynWr1WoIsKfpvCkqIUCaphECUQ9RDKUcwcCHMD9s6cVTk5ZkrUgNsg5+LkKaprmW4xhUUoqs69LSkkqlUm7rsKdjIfHccGOYMPQ7OzvBB3CWAFmPSqWiRqOh/f19tdttraysRLtzr1Fg3q46Q3CtjEAxHPB0jZSdJShl3XDJ347H2fkB8AFeAow3INOAonNenPe099SgE0b+e94w0isCUQKvYvOqQ0nhUf0YcoROUk6wee8ihyBlRqgYEiBYpPq4J+FGMUOB0cBwMGcoK++TJEnugFMQxXg8Vq/Xy3lmlI7afgqWCElAdvQCRNFQLFAQ0B8jSczuGQTmOUmSKAYDLfF+vnuwqFysJ6GGG2nkDIPNPXh/HE6/34+NQszBwsKClpeXJWVdpUkP8vsum4zL0+Q/cMRgkQfA00nZNlVSPJK0srISi5MkibrdrprNZkDH4XBydp3n6WnqgFWH0HMv4FV8WH33JC4EzkA7vMXTLSwshLHAsyNw7hXxekBbvCiGz3sHYghRnmImgHEyB5CdVPVVKtlWVubPc9sYEEgxnzfuz++glFLW9ouYmfG5AfNU3nA41NLSkg4PDyPeZt5dAeEfTk5Oot8f8T+pRzeIGMpioZOUnYngbeP4uZODHt6cnU3ON8ARSFlWCYJ0MBio3W7r9u3bQahSrFUuZ8erU1XYbDZzBVY4La9SRQ8ue10rI3BRKIACUuvth0FACHoxEfDTG3Q46YKVR7CAsxShYGQwRgg8Au0ZADcOGAHfZ48hY3HdOzv3gLHwKjJJASvhOXifIkrCc3Ph0fDaKBzGkFbt3E9SnOTjBCmEmpS1GfP23IQSflirj8vjcRSfd8E4wJCzo1NSzCHHiHklphO2HArrmQQu1gij6gbaeRwvMgONUH8COsR4odzMpxOI5XI59gnA3UxPT+vOnTtRYEVB2M7OTmwqOjg4CPRRbAN3FTUC0jUzAlLeEDgkkpTrA8geAlpIezoOxSamQ7FRCmJFNoEQEyOMXsmG53Z0Qbzv6UAvrgF+8lyU2gtkqFCjAg0jwDsAv92IeREO7+GcCZ4URCBlzVk9K8E+C+97T4MWOAgUDMGFeyG16pyBlHlPKVNiQi/W9SLj6e/p6U9CNNKaKCwOwNu3o6CulF6ezdic4HXo7fUahBUQsRSIkYUiTCBMdDQJp7K6uhpzu7y8rG63G2iOEIc5hTcADTgCuKp04bUyAkyA8wIOvyB0jo+PdevWrVxoMD09HXl2FtURAkLhqShnn6WJkqK8jKdY4VW0zkB0385KDpw6BeAp3ICTcezThyCkvBS+wtlxFII/KDuCjCfGQIGKELDxeHJGI3lsFARC0hn8fr8fRKSkSPNhLEhlIqzMnZOqrB1Vm4QlrLPvU2B81FNIiuPW8MyNRkNnZ2exbRyjjQFGFlg7DD1rxliZM1AOjoO5powYucPgYNQwjFyQlNLE6G5tbWllZSXmVVJwT4xDUhhZtiazVlddK3DtjADen8srzthphrLt7e3lNo7QSISY1yvLOHkWQ4BwE7sSqyLojMVbVLkBQPA9zVepTDoYIdxO7DB+L97BQPneBWoJHD5iWKTMQHo6FOE+PT3NNS2hIzPvTD0EnpEx4tWc6OJdgelJkkTJLAQhc4SXL/5dLmdn9uGleR/eBcPAPMKNgNbwoMw1FaEeEmKk8cyQcMwRyMR3NPIzDDuhHHE7aI6SceTMDQXr7BuSTk5O1O/39corr4RzkBRnYnY6neg1SDYBRIqsXCUKkK5ZnYATgUW2lBNcWVBgFUwyMauTQZIiBGCxSC3iCb3ib3FxMbrnwOZ7PI6CeArHqxm93txjeN4HyE5mwfekE5cSk6P0fAehnJqaiuwHTDSwmpz/wsJCKDWf01iVuSBkwlNJE6WECARmM//MlSMImH/uAwogE8A8upHDSBT3DmCMUHYIUBAFxsvRFdkc5tv3LRCqoLQQiCAAr2SED+B+zJvXHVxUzIPM8o4YXJAdiIK6Fd+rQkjFWvC+Hv76sy6TLrxWRsCJruLluwbhAzydx34BiK/Dw8MwCOSAPWZFSFB+/o9BQRBQZIScOF5SdNyhdsGFwCsLvUoPhh3y0ltaedxM0QwZBJQfqI+RmJuby8F/OhAxlx4Te5jEiTqgExTADS3vSyqLsAKjwxiZKxqPMG8UamGYUBR4ABTcU6tSvu8gCgJCYk691Je5Rul4R9AEXABjkxTw35/BPBC6ODdUnHMPJzA0GKPZ2Vnt7+/H56AVTiU6OTkJY0oTEuS1iAi4LoMMrpUR8DCgmClAUNJ0cvrQ4eFhFKcA90ejyWESCC6TiaHwTIBX6PkZdzyLAhKvt0fRUE5J8bOFhYXIXBD/IxggD6oPqR/390XJJMW7Opxm7EWlxHOMx+Ncuok55HsYMyrkKpVKpKnIpGAMSY36vVgfP4MBxcCgEZt7loHGIyAXEJZ7XdAT4YDH9igh78blFY+eJmU9GDfGyElJ3tk9vPcTYD74Du/laUgKzficcMRTg0+ePImDbXEMODDCKjdGzOVVhwTXihMoGgEuPA9epFSatG3Ck7BwVJchHOxxx8PzDM97U6PP+QUeU0sKtMCCAdkkhaARJnhu2a25F8K48gOvYeqdifYKNhAG5CH17JBhzhugVCixhyF4urm5OXW73fB6nmkAopNu8wyHl+ISpmAQ2OnHWnnpMhwO98Azgty81p/5YQ0YB8rq6UbuCVJDRqSMSPadm46QXBYwNBhMMik8zwlQ3pktwtwfRNZoNNTpdOL5BwcHgZbIuoBakGfnJ3hP14XLXs/TVOSOJu3G1ySlkn49TdN/mCRJQ9I/k/RxSR9I+qk0TfeTyQz+Q0l/TdJA0s+mafr/Xnagvjgec5dKWV94KYNxeHpgJpAc74vw03VIUu5nHnYAj1mUYkUeVt5/3/PDCAjC69+DhOS+ni3odruam5uLajv3RMS3JycnOTb/ooKkItHIu1FoBEGIQnmDFubbU4oYPlhx3pdOThBvZAiclMMI4rlBUigu42Y3IIipWOcAUmCcpHsxqo70nCCVsm5LzDX3I+Rjnn0+pcwo+HMxlKA7iEPe03kf39nKfPf7/WhkA0LxzsPIC2P0zNRleAC/nsecDCV9IU3TT0n6vKRfSJLkU5L+G0lfTtP0dUlfPv+/JP0HmrQVe12TRqK/diUjVb6ZghsEiDHf/YegAt08jcYCjsdZLzj+sPAutL7wknLPx7NJ2b57T8l5TQKeBtiOgBP7UxhCOOCxLwoCuchz6Jfn96Z6jmd73tpJQuYtTdOIOb2DEV6TikaISfdOxNPkwvG8PJuyXghaSWH0PNXraVQMFwpGLI8S0L7MU7dAbp7nYR7r5ErrhUkYAO7J++LZpaxJqRt0DCPryBqxWUjKkAljRiaRS4ybp42d30DG3bBdpQGQnq+z0IakjfN/HyRJ8pakW5L+uiZtxyTpf5P0f0v6r88//yfpZFa/miRJPUmSG+f3uZLLJ4HFxggQL7sRYOFJJ0Hq0MyBIhcvQZXyTTwlhSLhhRwWeukpiuBGBfjK32wcQXC9CKmoMB4D+ngIWRzWMxY3PkWvRUyOMjBvxbgaHkRSVMdJym3PxcB6ztuRGBe/K2VNOhza4vFAKFwgPD7DoPCerBHhh4c2rIFXdYIMeB7j5NkYZN8FyD1YR4wf9Q3cy426lNUH0J+C9eDeKPbBwUFknlh7MivFzIM7kb8QTiCZHELyb0j6mqQ1U+xNTcIFaWIgHtmvPT7/7NJG4MPIEJSOSfXCFvfm1LqTogH2ekmqhwMgCmCvj4MYnu+hdFJWI4+X9WwD9/HyVjdWzmv4dmEXXt6Z7/MslNLrBhDM0WgUIRGGCKPkAuXkJ3CU5zv0lZQzAgg4ZBjHd5PudJ4DQpVnwgUwRyidZwUgIfl9jLBXCTIeJ0O5L892pp7LSVSchMuTIzcpOyYdAyplvSxYz9FopGazGfMpKXgodyLj8Tia3YAOTk9P1etNuvSDXJl/ZMpJ8csag+c2AkmSLGrSP/C/TNO051Y+TdM0ec624Xa/7/rcARSh6GEQqGJcjAUFIgOBh8NhpPTwqsRxLhx4OIfajgActjrRh+CgoA43EVgp85i810X5ZucQeDcnEBFcNzgoZrGakDgUgcXgePrKDYsbORSrqHxAeydFq9VqcAXuUZ1848K7ArcxCLwrc+AxPMrtv4dRxvh545eiJ/dw0edRUhg5Qg14Euc++D/ozVENcyQpF0rg2fmOI87j4+PoFsWeBE7PwgjgZJARn9PLXs9lBJIkqWhiAP5pmqb/5/nHW8D8JEluSGqff/5E0h379dvnn+Wu9Ls4d8AV4vx342csTFHhEGi/+C65ckkhvCw8AoKVdtKKhSCmlhTMvTPoCDCC5qW1rkwueG7hvX04wo8y+bu7UcT4MG43RigyhsHZdt4J4pPn+Z4GKuOYG1J6ZBCczCPEWF5e1ng8jr75GI2L1pF7+jiYM2fmMYggLQ+vuDfvjdwQcjmZyD19vfiMZzryc+VlHZ0L4buMOU2zNHGx6MzXjNBvOJwcYEJJ+8nJiQaDgaQsHPJ6BSeyXUde1CB8R2LwnO3/DUlvpWn6D+xHvyvpZ87//TOS/oV9/tPJ5Pq8pO5l+YAPezksP14AK18UEj5zksyLPbgPgopAONRl8cjVelrQK9TwLI5YivEb4YETZC54IA7eAfJT0jPC5LCQON89Lf+GPHSjxM8RZISKuBSvwxw5uSllfAXGA+PinpR7F6v1UFbP5kDKsS+Brc0+t5JyrbjhOMjBezYG4+GpP+ab32e9/HM4E+RHyjo+wRfwDBClG2yq/1hfuArekXlBFijZhtfwwjMPbRhrcW2/nY48z/U8SODfkvSfSnozSZJ/ff7ZfyvpS5J+O0mSvy3pgSYHk0rSv9QkPXhXkxThz73w6J7jInb0SXDo6Q0/8NgOiT3exnsXm43ghYqpSF8UFtf/9p+5R8bQ4PEc6hFDS4rOP3g6FN0zFYyhaBRALpB+jgScZZaycxEcVbiBY4zE5UmS3wDkBBahFnl5J9lQHIyKp+OKXs17/zPGi+aTcTrC8opAf4bvHOQ5RQPHWPieIyI35sgHc8k9K5VKtKEr7hlw2eR9Wb+TkxN1Op0oQuN9MXQYtCIf8H3hBNI0/UNJH5aP+LELvp9K+oVLjepDLn9p/hRz7w6r+YMxYJMH8BUL6/8mNeQ7Ai+K1/FwKJDHtB6zn89JDp7jdYbDYQ52Ysw8/YS3pbAFg+T5YikTUC/iQYCYtyJ05nMnHKV8SOU5av5N+pR7OCz37jrsKCSMQLAxrm68PPb1uQDB+NxjNKmXcMVl/LD8HsYxftbaPWlxLbg8xGI9fE44Y5A5gOXnd12ekEWMOIVfjIMzMrxWhJOyWSeec5WcwLUqG+Zy4XUPUYytPa3EDjovyHGo5fwBgsrPqOd35MB9HXIiaEVP68wuP3OWHLTBuEEuCBvv5Cw98+DGwAk8j6OL8+be1dNoHmpgzBzZYIR4J383SRHfcrmy+PtiUD3F6t4NSM67UFF3eHgYFZ94bzcwKBvPcsPH+7ohoAIT+I1Bck/MnBRRgKMHNkvxTM8isHYYJk8pY/BZ14ODA3U6ndgHA4q5qNvwVaYIr70RcHaeunQWBCGj/p/PfQcgvw+vwEXpLd7U/3b2n3gQxXVOolgo5PGllKUS8YT8vnsHKSOmGI8/m+/xfAwcAsI8FMMW97qeW0fgnTRkzFNTU7nNMZQt8weuwXdy8jOPiXk/1s0NFWPDs/Nv944U84ASeC/GC5HqfAuywNy5YoMC+H3mxkM65Ib7eAEUv+/3Ozo6CofD/BOaYlhAeRiK/f39OLDUG9ywth5+FGXkMte12jvAy7uAuxBRn02ZLQvQ7/dzMdni4mIIsC+yx2t4ZSq/WDAXaD8ZmMtjWymfWnKv4KGB39N5Au7NtlM+K4YMPN/Tj5CYSZJEPYTnmT0mpeTZycFiMQ/1En6sthfnsE0bz8gYeJYTcf6Zx++OQBBw0Azcj9dclMvlOOfPEQvGiDFKmaErEpeMxXkkh/GMhRy+zy8bwhwdueH0bBCKTJgCGmE8EKGl0qSt+8LCQii+Iww3ot83TuCjdhUZUZ8AFgkyTJoIs8epDsHw4CwEFXIOIX3Ci8paHIMbEClrk1Ukr9ybOzEFEuCeTtx5/El86dkLPLePmTG4p+X/zqe4N3OPjRHhM8qW8U6sh3MrknIK5HPB3/6+jME5Bl9fDAXPYG19rp1xh5DzkI91o/jI19KNYnGuiqSkZ41oD+ZVkqPR5DBTyElvgAIxSXhJeFMqlcIASFkqkHMfx+NJY1XqW/ydi/L/ote1DAek/BkELAyxPAvMYgHbXCjp4edIwL2De+uLUoZStk/dKwGlTPld0Pw5XrQi5Xvu8Tt4Do/TeT8PA/BwXCiN71Yk1HCDhrFB+b0DD+OFA/AdjGzlRZH5XZQAD3d8fKxer5d7DuNxA+eG0oXbU5HOjPOOUr4giOrEYtjB/T4s1ve5cwPkcw868LCB9OXU1FSuh8Lp6Wl0DMZouVEmUzMzMxMNXTGabLYaDAaqVqthSDD8LkfF8V7mujZIoJgZKH7mjTs8peNbh51Brlar0SbM+90Dj+ENXECLLaSkZ3sIenbAsxAuhDDEwFiUEtZ8amoq9jMQHrgHKcJB3xDDmBwpeNmuk4UgnvE466xDPI1gUkjlra5BTqRTMZD09PfYlfu4t8bQYdBQ3uLvupHmfXwuMYx++Aj3A2kUY3w3Qqw18b3n//25yBJrzK5U1ph1c2/O3HpTGAhgwoM0TXOhK8aL6tU0nWzsIhwpVhy6MbvMdW2MgHuJokVHiWBj2fNOrtnDAGfsWUgW8/T0NDrPYDjc07CoKJd7CeCyCxE5/2Is7DDVhRougLp7hA4I7ox5EWZLyu1nwPs6AUV6j4rEIsPtRVQYH5Qd48pGI/7vRUBSdkqz798nA+LQFcFnLfiZ5/GLKVdQCVWFzg0QLnn1nnthjJHn7h1d+DsUjYWvFT9nLiWFoaT1GrJA6a+HTy4n3qsSY0ZtAvNGtSZGj7VgDq/iujZGwK8iM4oXYWFQSu9Jj6eXsnQX7bHd0zqT7p7UC2a8Ead7f4e2eA5P/yGsnjVw7gFPBAPPd0lRFvPl7Nv3lCLhCXPiXpDegs4FoBgoAO9XDEXce/IMoCrK4GvjvRFBIo5cipyF5+899CkKOmjB05u8o8fizBf39OwMBgh4zzuxfkVE58gvTdMIQ/iMtvYgT9bm9PRUS0tLsX6cKQFRi6M5PDzMkaf8G2MHikDeeAfGc9nr2hiBYjjg8SFbOuk+C8wCvqMACKfDbgyEp29QTlc6PnOPIWVpQC/rlfI5acYvZfF/kiTRZMIFD2MDiqAGH4VwkpEwwNNkXO4d4Qm89yIK7LDWDxllHri3NzPFMKFEzCnvsLCwkDsZGOThHZwcETFvHuK4h2QemQNn4JlrYDohHYU3viuR72B0XLmRI57jyI21xPBhyBxJwAUgJ/ReODw8DGSJo4Ez8NACDoG5cYOO0fKsh4d/l72uhRFw0siFx8MABAeC6OTkJFqDQcaUSpPyTO8whACfnp6qVquF8rB9GMMATEuSJMqPeb6z/aAHj/NQDk+4K11BAAAKbUlEQVQLpWkasJ/v+7ZR4DPf96PRSNUVMwuOVi4q6sGoUdzC80BPxNR+yCZxLh5Qyhhs0lhpmkZbM9aBzTO+dlK2F6IYezvXIWW1FJ41AVmggE4M8ns+RowdoRTfY66do+EdixwAGRnWgjkm1GBNyO8Xm4/0+30lSRIt7XkW/QgIX8kUeOGTk6mgVOdMeGcPD1/kuhZGwBcGBeJzFhYlR6BhsoFiKLSz5lNTk6OnDw4OwhhIWTxJ7b4/kxCD5yOQbgy8SSRjdGF3hfDiIj5zss3zxCAYyC9JuUYqjNO9KHDfMwGennOYyxi9gYhXFOKJYbaHw8lZjv1+P4wTP5+eno5DQLxk2D09CoTCoKiODqTsiDDnCBweMz5ORIZMY65BVTzHURre3DNLTiKyZh4esBbcl/HQl6KYmdjb24u5wph6/4pimCBlZyAQnrEN3nWCd/HM1Itc18IISPnWYv6HzxCy6elp7ezsBFPt+85RVGAtgseCgBAQTGJuFo7fAY55PO7eDYODN/ENTt6zANb3ovJZnjs3NxdewhEDPAVIhT+ucI6cpHxTVCe8nHRCsCjCwWgSutBslbFD+i0sLKher0dD0ampKbVaLfV6vZhHEBRIzZ/Pu6PUfOa8AvvtPWULqmEtMJSgLEnRyANlxnj7nHja0cliL9witMKoFg0WRpL7VSoVdTqdWF9+hiEGNYIGer1eoD52H0LEsm6saTEjcBk0cG2MgPRsWzH+xghIk8WgRNi9DYviZJk0sbic/iJlHWL8Xn5Kjse/xXF52amTO5JyHheB8P3nTgQCNxFmFAhhd8/uRUQYJ493UUjPTHiJsxOSvBt8RdEg1uv1CANg4vGIUsZ3MCcYKbw0kBaUhZGU8h2NeE8nVr0EnHWB7OVd6WjE3LkcsAaEFTiIotf3MM5DT9YM0g8PTQVgsUuVpzBHo1Ec3kpo6qcqwRPw/KmpqeBU4DK4n4+1qAcvel07I+AXXqtIkqBUUtYJ2IkuFABlYA+3TyhKPT09HV5RUqQIiyfiIswooy+KZwakzMt7us9ZbBSIzASpySRJ4vw/iDzQCyk5oDHPdchP5x/KkBFE3wADi+3IBsHr9XohlH4UmFcSYqyAx6wDoQKogHnwPg4OoT2Vx7s4T4EB5feOjo60v78f28Cpa8BIYkBQLj9Jqej5UXDOrcD7e/m1lPf+nq0gM+VhAQYyTdM4Js27OoMGIF1ZW+/94MVY7hAve10rIyDlwwImHmuLx3YyDuvv2QD/GcrNgRB4HVI3nhpzNOFVfsTlxdQizLRnDjyX70bAq9HcSzNeKYPLIAV+RjaBIh+MiBsasiaESSike0UMCaEPyu/EqnMbGBO644K4nMR1OM0xZ5JyXpj/u+HAWGDUWFNJuWPaOYCWmLpYx+C/T0dl7kNoAc8Ce+/Gj++ORv9fe2cTYlUZxvHfg6RIM5fbjCliUhpuXNUg4kJcVrqZ2rnKRdCmoBYtDDduC3IRRFAkWERuKnIT9EHQKsvCz8SPSigxLZKZuMxHo0+Lc/7nPvc2F8dm9L2H+/7hcs+ce7nzP+f5eJ/neZ/3PTeqsFzRSezDEGfNmkjW8bFsk5OTjIyMVIuDZODK+YeGhjqetCzHJB2Ocom2MBCFwfkQPbhukgp5yjehHT1IQeM8tkJ1oBpBYhEohogySGgXrGJ+HQ1AhqVpx9hEJKcio1WzjYwtbi2m9fhSTv0vKYEUQ05JBgx0PKVHjmL16tXVaNNsNivF10gs5Vu5ciXNZrNjH0Ypaow+pNzu7acSyRDFVUU6s87+hVhfiTvxxqKf0p0466Lqu6IjrbeXI47NNDLaGEJHOWlUjYYuJxSNLU4LSr+izCQL1XekP0pNonNWdNdoNKraSNwBG9oLwcRbOhlT26UwfqGWTiAaP3QaqXq6dUO1u3AchaODkLeVcSqvk3GqiBSr/1KguGGnlPjmzfYmmNFTxwU3sbijEUHOI26nHaeK5MCUWyufltLFFlz9j9jpp9+XsaqbUgoedxdWtV/3S6P8xMQEw8PD1TXFkV3z4XEdgaKHubm5avSTk4hGqO8rmovFSqVZMrq5ubkqqpEzaLVatFqtKkWRs4kdhJJfbG2G9mavcsAyPLUya7WpRn6lfnGgUDoFVE5TnOVA1RgknVMkJ7mpftRoNLh+/XpHT4J0Pk4xx3RtsY6gVguIdNMlUI22sdU0dpPpPVZn3b2aNlQoqBs9X7MItOeeZSBxd19Vu6G9jFXFKzmP7lAdOjfgiIVNORSFstPT04yOjlbKqFFUiqsUIhYrYwQTX6rUx3ZqhfLxOjSnHe+BwnMppu7zzMwMjUaj47mFUujYxKQRT5+rsUdFNDUzReOK3ZvQWVTV9Wu0lWx0H1Q/kOFq0IhNPro2OUwZ5NTUVMeeBd2zBzGC0ixNdFjT09NMTExU8lOqo8Lr1NRU1VCkezU7O1tt1y6n091XEhvLNIgsRSRgS/EjiyZh9gfQAv5MzWURWEW9+UP9r6Hu/OHOXsOD7n5/98m+cAIAZnbM3bek5vF/UXf+UP9rqDt/SHMNtUoHMjIylh7ZCWRkDDj6yQm8lZrAIlF3/lD/a6g7f0hwDX1TE8jIyEiDfooEMjIyEiC5EzCzJ8zsnJldNLO9qfksFGZ2ycxOmdlxMztWnhsxs8/N7EL5fl9qnhFmdtDMrpnZ6XBuXs5W4PVSLifNbCwd84rrfPz3m9nlUg7HzWxX+Ozlkv85M3s8Des2zGy9mX1lZj+a2Rkze6E8n1YG3Yso7uYLWAb8BGwElgMngM0pOd0G90vAqq5zrwJ7y+O9wCupeXbx2wGMAadvxZnieZKfUjyCbhtwtE/57wdemue7m0t9WgFsKPVsWWL+a4Gx8ngYOF/yTCqD1JHAVuCiu//s7rPAYWA8MafFYBw4VB4fAp5MyOU/cPevgb+6TvfiPA686wW+AZpWPII+GXrw74Vx4LC7z7j7LxQPyN16x8gtAO5+xd1/KI//Bs4C60gsg9ROYB3wa/j7t/JcHeDAZ2b2vZk9W55b4+3HsP8OrElD7bbQi3OdZPN8GS4fDClYX/M3s4eAR4GjJJZBaidQZ2x39zFgJ/Ccme2IH3oRz9Vq6qWOnIE3gYeBR4ArwGtp6dwaZjYEfAi86O6T8bMUMkjtBC4D68PfD5Tn+h7ufrl8vwZ8TBFqXlW4Vr5fS8dwwejFuRaycfer7n7D3W8Cb9MO+fuSv5ndQ+EA3nf3j8rTSWWQ2gl8B2wysw1mthzYDRxJzOmWMLN7zWxYx8BjwGkK7nvKr+0BPknD8LbQi/MR4OmyQr0NmAgha9+gK0d+ikIOUPDfbWYrzGwDsAn49m7zi7Bieec7wFl3PxA+SiuDlNXSUAE9T1G93ZeazwI5b6SoPJ8Azog3MAp8CVwAvgBGUnPt4v0BRcj8D0V++UwvzhQV6TdKuZwCtvQp//dKfidLo1kbvr+v5H8O2NkH/LdThPongePla1dqGeSOwYyMAUfqdCAjIyMxshPIyBhwZCeQkTHgyE4gI2PAkZ1ARsaAIzuBjIwBR3YCGRkDjuwEMjIGHP8CyEburYW8U6QAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6K8ptTLYGOrv",
        "colab_type": "text"
      },
      "source": [
        "## Data Augmentation\n",
        "Una vez tenemos el conjunto preprocesado y con la estructura de directorios adecuada, procedemos a la aumentación de datos.\n",
        "En caso de pretender reproducir los datos, puede obtener el conjunto debidamente estructurado del siguiente enlace https://drive.google.com/drive/folders/1KG8aschyKw5VvmOffGCN9UaNwPn043QJ?usp=sharing, y modificar la ruta de cada partición devidamente."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZI-B6leSU1MT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/train\"\n",
        "validation_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/validation\"\n",
        "test_path = \"/content/drive/My Drive/SIGE DATASETS/DataSet_Covid_Normal_Xray_Classification/validation\""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lq1hRXwOGTa-",
        "colab_type": "code",
        "outputId": "6bb8e5e1-abca-4762-b325-c0cd68929ce4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "datagen_train = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1./255, \n",
        "                                                          rotation_range=45, \n",
        "                                                          zoom_range=0.2,\n",
        "                                                          width_shift_range=0.2, \n",
        "                                                          height_shift_range=0.2, \n",
        "                                                          shear_range=0.2,\n",
        "                                                          horizontal_flip=True, \n",
        "                                                          fill_mode=\"nearest\")\n",
        "\n",
        "train_augmented_data = datagen_train.flow_from_directory(\n",
        "    train_path,\n",
        "    target_size=(224, 224),\n",
        "    color_mode=\"rgb\",\n",
        "    classes=None,\n",
        "    class_mode=\"categorical\",\n",
        "    batch_size=32,\n",
        "    shuffle=True,\n",
        "    seed=12345,\n",
        "    follow_links=False,\n",
        "    subset=None,\n",
        "    interpolation=\"nearest\",\n",
        ")"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 236 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g9l6iWx_WugS",
        "colab_type": "code",
        "outputId": "9a5e23fe-41c7-4589-a30c-232d54f16b0e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "datagen_val = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "validation_data = datagen_val.flow_from_directory(\n",
        "    validation_path,\n",
        "    target_size=(224, 224),\n",
        "    color_mode=\"rgb\",\n",
        "    classes=None,\n",
        "    class_mode=\"categorical\",\n",
        "    batch_size=10,\n",
        "    shuffle=False,\n",
        "    seed=12345,\n",
        "    follow_links=False,\n",
        "    subset=None,\n",
        "    interpolation=\"nearest\",\n",
        ")"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 80 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L7er4LKY5Z51",
        "colab_type": "code",
        "outputId": "5fd99672-1df2-4da3-d90e-b63435c6e766",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "test_data = datagen_val.flow_from_directory(\n",
        "    test_path,\n",
        "    target_size=(224, 224),\n",
        "    color_mode=\"rgb\",\n",
        "    classes=None,\n",
        "    class_mode=\"categorical\",\n",
        "    batch_size=10,\n",
        "    shuffle=False,\n",
        "    seed=12345,\n",
        "    follow_links=False,\n",
        "    subset=None,\n",
        "    interpolation=\"nearest\",\n",
        ")"
      ],
      "execution_count": 52,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 80 images belonging to 2 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-jfIvsI3cZcX",
        "colab_type": "text"
      },
      "source": [
        "## Aproximación CNN inicial"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XQmRJ1cacYKv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.optimizers import Nadam\n",
        "from tensorflow.keras import Sequential\n",
        "from tensorflow.keras.layers import Dense, Conv2D, Flatten, BatchNormalization, MaxPooling2D, Dropout, AveragePooling2D\n",
        "from sklearn.metrics import classification_report, confusion_matrix"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v8_kU7RGfD_w",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#create model\n",
        "model = Sequential()\n",
        "#add model layers\n",
        "# conv + relu + conv + relu + pool\n",
        "model.add(Conv2D(64, kernel_size=(7,7), padding='same', activation='relu', input_shape=(224,224,3)))\n",
        "model.add(Conv2D(64, kernel_size=(3,3), padding='same', activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(rate=0.25))\n",
        "# conv + relu + conv + relu + pool\n",
        "model.add(Conv2D(64, kernel_size=(7,7), padding='same', activation='relu'))\n",
        "model.add(Conv2D(64, kernel_size=(3,3), padding='same', activation='relu'))\n",
        "model.add(AveragePooling2D(pool_size=(2,2)))\n",
        "# conv + relu + conv + relu + pool\n",
        "model.add(Conv2D(32, kernel_size=(7,7), padding='same', activation='relu'))\n",
        "model.add(Conv2D(32, kernel_size=(3,3), padding='same', activation='relu'))\n",
        "model.add(Conv2D(32, kernel_size=(3,3), padding='same', activation='relu'))\n",
        "model.add(AveragePooling2D(pool_size=(2,2)))\n",
        "# fc + relu\n",
        "model.add(Flatten())\n",
        "model.add(Dense(64, activation='relu'))\n",
        "model.add(BatchNormalization(epsilon=0.01))\n",
        "model.add(Dropout(rate=0.5))\n",
        "model.add(Dense(2, activation='softmax'))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3gh-xB1ChQk9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 697
        },
        "outputId": "83c9a5db-078e-4f6b-b06a-a2e64940e4a0"
      },
      "source": [
        "# https://stackoverflow.com/questions/41665799/keras-model-summary-object-to-string\n",
        "stringlist = []\n",
        "model.summary(print_fn=lambda x: stringlist.append(x))\n",
        "short_model_summary = \"\\n\".join(stringlist)\n",
        "print(short_model_summary)"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_232 (Conv2D)          (None, 224, 224, 64)      9472      \n",
            "_________________________________________________________________\n",
            "conv2d_233 (Conv2D)          (None, 224, 224, 64)      36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_5 (MaxPooling2 (None, 112, 112, 64)      0         \n",
            "_________________________________________________________________\n",
            "dropout_5 (Dropout)          (None, 112, 112, 64)      0         \n",
            "_________________________________________________________________\n",
            "conv2d_234 (Conv2D)          (None, 112, 112, 64)      200768    \n",
            "_________________________________________________________________\n",
            "conv2d_235 (Conv2D)          (None, 112, 112, 64)      36928     \n",
            "_________________________________________________________________\n",
            "average_pooling2d_6 (Average (None, 56, 56, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_236 (Conv2D)          (None, 56, 56, 32)        100384    \n",
            "_________________________________________________________________\n",
            "conv2d_237 (Conv2D)          (None, 56, 56, 32)        9248      \n",
            "_________________________________________________________________\n",
            "conv2d_238 (Conv2D)          (None, 56, 56, 32)        9248      \n",
            "_________________________________________________________________\n",
            "average_pooling2d_7 (Average (None, 28, 28, 32)        0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 25088)             0         \n",
            "_________________________________________________________________\n",
            "dense_8 (Dense)              (None, 64)                1605696   \n",
            "_________________________________________________________________\n",
            "batch_normalization_204 (Bat (None, 64)                256       \n",
            "_________________________________________________________________\n",
            "dropout_6 (Dropout)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_9 (Dense)              (None, 2)                 130       \n",
            "=================================================================\n",
            "Total params: 2,009,058\n",
            "Trainable params: 2,008,930\n",
            "Non-trainable params: 128\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DJCipmrEmDo0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(loss=\"categorical_crossentropy\", \n",
        "              optimizer=Nadam(),\n",
        "              metrics=[\"accuracy\"])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BpbRnNmSmhjM",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "7732ed75-59b8-49ba-c433-ca973e7835bb"
      },
      "source": [
        "H1 = model.fit(train_augmented_data, \n",
        "                        steps_per_epoch=8,\n",
        "                        validation_data=validation_data,\n",
        "                        validation_steps=8,\n",
        "                        epochs=50)"
      ],
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.7039 - accuracy: 0.6780 - val_loss: 0.5995 - val_accuracy: 0.5375\n",
            "Epoch 2/50\n",
            "8/8 [==============================] - 4s 482ms/step - loss: 0.5520 - accuracy: 0.7797 - val_loss: 0.4493 - val_accuracy: 0.9125\n",
            "Epoch 3/50\n",
            "8/8 [==============================] - 4s 444ms/step - loss: 0.5069 - accuracy: 0.8051 - val_loss: 0.3119 - val_accuracy: 0.8250\n",
            "Epoch 4/50\n",
            "8/8 [==============================] - 4s 438ms/step - loss: 0.4750 - accuracy: 0.8220 - val_loss: 0.5347 - val_accuracy: 0.6250\n",
            "Epoch 5/50\n",
            "8/8 [==============================] - 4s 446ms/step - loss: 0.3397 - accuracy: 0.8432 - val_loss: 0.2591 - val_accuracy: 0.8750\n",
            "Epoch 6/50\n",
            "8/8 [==============================] - 4s 446ms/step - loss: 0.4350 - accuracy: 0.8178 - val_loss: 0.8801 - val_accuracy: 0.6750\n",
            "Epoch 7/50\n",
            "8/8 [==============================] - 4s 442ms/step - loss: 0.4365 - accuracy: 0.8305 - val_loss: 0.4598 - val_accuracy: 0.8750\n",
            "Epoch 8/50\n",
            "8/8 [==============================] - 4s 479ms/step - loss: 0.3669 - accuracy: 0.8771 - val_loss: 0.4004 - val_accuracy: 0.8500\n",
            "Epoch 9/50\n",
            "8/8 [==============================] - 3s 428ms/step - loss: 0.3642 - accuracy: 0.8686 - val_loss: 1.0965 - val_accuracy: 0.6625\n",
            "Epoch 10/50\n",
            "8/8 [==============================] - 3s 437ms/step - loss: 0.3524 - accuracy: 0.8432 - val_loss: 1.4431 - val_accuracy: 0.5750\n",
            "Epoch 11/50\n",
            "8/8 [==============================] - 4s 443ms/step - loss: 0.3097 - accuracy: 0.8941 - val_loss: 0.3710 - val_accuracy: 0.8750\n",
            "Epoch 12/50\n",
            "8/8 [==============================] - 4s 449ms/step - loss: 0.3255 - accuracy: 0.8602 - val_loss: 1.3582 - val_accuracy: 0.5000\n",
            "Epoch 13/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.3685 - accuracy: 0.8517 - val_loss: 0.9493 - val_accuracy: 0.5250\n",
            "Epoch 14/50\n",
            "8/8 [==============================] - 4s 449ms/step - loss: 0.3745 - accuracy: 0.8432 - val_loss: 2.8796 - val_accuracy: 0.5000\n",
            "Epoch 15/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.3270 - accuracy: 0.8771 - val_loss: 2.1406 - val_accuracy: 0.5000\n",
            "Epoch 16/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.3152 - accuracy: 0.8941 - val_loss: 0.1984 - val_accuracy: 0.9500\n",
            "Epoch 17/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3328 - accuracy: 0.8432 - val_loss: 0.1170 - val_accuracy: 0.9750\n",
            "Epoch 18/50\n",
            "8/8 [==============================] - 3s 437ms/step - loss: 0.3032 - accuracy: 0.9153 - val_loss: 0.2383 - val_accuracy: 0.9500\n",
            "Epoch 19/50\n",
            "8/8 [==============================] - 3s 437ms/step - loss: 0.3144 - accuracy: 0.8856 - val_loss: 0.2397 - val_accuracy: 0.9250\n",
            "Epoch 20/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.3043 - accuracy: 0.9025 - val_loss: 0.8835 - val_accuracy: 0.5625\n",
            "Epoch 21/50\n",
            "8/8 [==============================] - 4s 443ms/step - loss: 0.3813 - accuracy: 0.8559 - val_loss: 0.1507 - val_accuracy: 0.9625\n",
            "Epoch 22/50\n",
            "8/8 [==============================] - 4s 486ms/step - loss: 0.3242 - accuracy: 0.8814 - val_loss: 0.2980 - val_accuracy: 0.9375\n",
            "Epoch 23/50\n",
            "8/8 [==============================] - 4s 483ms/step - loss: 0.3735 - accuracy: 0.8814 - val_loss: 0.7596 - val_accuracy: 0.6125\n",
            "Epoch 24/50\n",
            "8/8 [==============================] - 4s 441ms/step - loss: 0.2531 - accuracy: 0.9110 - val_loss: 0.1340 - val_accuracy: 0.9750\n",
            "Epoch 25/50\n",
            "8/8 [==============================] - 4s 442ms/step - loss: 0.3014 - accuracy: 0.9025 - val_loss: 0.2037 - val_accuracy: 0.9125\n",
            "Epoch 26/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3455 - accuracy: 0.8983 - val_loss: 0.1542 - val_accuracy: 0.9625\n",
            "Epoch 27/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.4703 - accuracy: 0.8517 - val_loss: 2.3893 - val_accuracy: 0.5000\n",
            "Epoch 28/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.3764 - accuracy: 0.8602 - val_loss: 1.6356 - val_accuracy: 0.5250\n",
            "Epoch 29/50\n",
            "8/8 [==============================] - 3s 425ms/step - loss: 0.3236 - accuracy: 0.8898 - val_loss: 0.8488 - val_accuracy: 0.7125\n",
            "Epoch 30/50\n",
            "8/8 [==============================] - 3s 431ms/step - loss: 0.2635 - accuracy: 0.8771 - val_loss: 0.3592 - val_accuracy: 0.9125\n",
            "Epoch 31/50\n",
            "8/8 [==============================] - 4s 438ms/step - loss: 0.3191 - accuracy: 0.8729 - val_loss: 0.4564 - val_accuracy: 0.9250\n",
            "Epoch 32/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3334 - accuracy: 0.8771 - val_loss: 0.4202 - val_accuracy: 0.8000\n",
            "Epoch 33/50\n",
            "8/8 [==============================] - 3s 435ms/step - loss: 0.3292 - accuracy: 0.8856 - val_loss: 0.3806 - val_accuracy: 0.8750\n",
            "Epoch 34/50\n",
            "8/8 [==============================] - 4s 490ms/step - loss: 0.2529 - accuracy: 0.8983 - val_loss: 0.2398 - val_accuracy: 0.9500\n",
            "Epoch 35/50\n",
            "8/8 [==============================] - 4s 446ms/step - loss: 0.3415 - accuracy: 0.8983 - val_loss: 1.2985 - val_accuracy: 0.5375\n",
            "Epoch 36/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.2904 - accuracy: 0.8814 - val_loss: 0.2095 - val_accuracy: 0.9625\n",
            "Epoch 37/50\n",
            "8/8 [==============================] - 3s 423ms/step - loss: 0.2712 - accuracy: 0.8941 - val_loss: 0.5018 - val_accuracy: 0.7375\n",
            "Epoch 38/50\n",
            "8/8 [==============================] - 3s 435ms/step - loss: 0.3030 - accuracy: 0.8814 - val_loss: 0.2421 - val_accuracy: 0.9500\n",
            "Epoch 39/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.3304 - accuracy: 0.8347 - val_loss: 0.3642 - val_accuracy: 0.8750\n",
            "Epoch 40/50\n",
            "8/8 [==============================] - 3s 432ms/step - loss: 0.2677 - accuracy: 0.9110 - val_loss: 0.4839 - val_accuracy: 0.7625\n",
            "Epoch 41/50\n",
            "8/8 [==============================] - 3s 429ms/step - loss: 0.2384 - accuracy: 0.9110 - val_loss: 1.0481 - val_accuracy: 0.5625\n",
            "Epoch 42/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3383 - accuracy: 0.8729 - val_loss: 11.0092 - val_accuracy: 0.5750\n",
            "Epoch 43/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.6276 - accuracy: 0.7669 - val_loss: 4.3787 - val_accuracy: 0.5000\n",
            "Epoch 44/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.6055 - accuracy: 0.7712 - val_loss: 0.6729 - val_accuracy: 0.8250\n",
            "Epoch 45/50\n",
            "8/8 [==============================] - 4s 446ms/step - loss: 0.3828 - accuracy: 0.8390 - val_loss: 1.0511 - val_accuracy: 0.5250\n",
            "Epoch 46/50\n",
            "8/8 [==============================] - 4s 442ms/step - loss: 0.3916 - accuracy: 0.8263 - val_loss: 0.2620 - val_accuracy: 0.9625\n",
            "Epoch 47/50\n",
            "8/8 [==============================] - 3s 437ms/step - loss: 0.3093 - accuracy: 0.8644 - val_loss: 0.4744 - val_accuracy: 0.8500\n",
            "Epoch 48/50\n",
            "8/8 [==============================] - 4s 485ms/step - loss: 0.2777 - accuracy: 0.8898 - val_loss: 1.1862 - val_accuracy: 0.5875\n",
            "Epoch 49/50\n",
            "8/8 [==============================] - 4s 441ms/step - loss: 0.2486 - accuracy: 0.8983 - val_loss: 0.9525 - val_accuracy: 0.6375\n",
            "Epoch 50/50\n",
            "8/8 [==============================] - 4s 481ms/step - loss: 0.3037 - accuracy: 0.8814 - val_loss: 0.4232 - val_accuracy: 0.8500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N8LO8CUhrBS7",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "outputId": "0e352016-dbee-47bd-a8bf-0a68be5fba8a"
      },
      "source": [
        "# Visualizamos valor de accuracy de entrenamiento\n",
        "plt.plot(H1.history['accuracy'], label='accuracy')\n",
        "plt.plot(H1.history['val_accuracy'], label = 'val_accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0., 1])\n",
        "plt.legend(loc='lower right')"
      ],
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f9908f9fe10>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 58
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aerXFwKASAox",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "outputId": "8aa2fb20-d78f-434c-fbfd-5033a0da5efa"
      },
      "source": [
        "# Visualizamos valor de pérdida de entrenamiento\n",
        "plt.plot(H1.history['loss'], label='loss')\n",
        "plt.plot(H1.history['val_loss'], label = 'val_loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('loss')\n",
        "plt.ylim([0., 1])\n",
        "plt.legend(loc='upper right')"
      ],
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f9908f13400>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 59
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0EgVR12JTAe5",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "fc6d4900-f1aa-4755-ee7b-54ad525da737"
      },
      "source": [
        "# Realizamos las predicciones sobre el conjunto de test, y \n",
        "# mostramos los resultados de la matriz de consufión, además\n",
        "# de diversas métricas.\n",
        "predictions = model.predict(test_data,\n",
        "                     batch_size=10,\n",
        "                     verbose=1)\n",
        "y_pred = np.argmax(predictions, axis=1)\n",
        "print(confusion_matrix(test_data.classes, y_pred))\n",
        "print(classification_report(test_data.classes, y_pred, target_names = ['covid', 'healthy']))"
      ],
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "8/8 [==============================] - 0s 33ms/step\n",
            "[[28 12]\n",
            " [ 0 40]]\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "       covid       1.00      0.70      0.82        40\n",
            "     healthy       0.77      1.00      0.87        40\n",
            "\n",
            "    accuracy                           0.85        80\n",
            "   macro avg       0.88      0.85      0.85        80\n",
            "weighted avg       0.88      0.85      0.85        80\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qnYFbRVUVYGZ",
        "colab_type": "text"
      },
      "source": [
        "## Transfer-Learning\n",
        "Para tratar de mejorar los resultados del anterior modelo, usaremos la técnica de transferencia de aprendizaje, empleando diversos modelos pre-entrenados para tratar de obtener el mejor resultado de clasificación posible."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FNqC2inx4uaI",
        "colab_type": "text"
      },
      "source": [
        "### VGG16 Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sAK2_kUddMR_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.applications import VGG16\n",
        "from tensorflow.keras.layers import AveragePooling2D\n",
        "from tensorflow.keras.layers import Dropout\n",
        "from tensorflow.keras.layers import Flatten\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.layers import Input\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.optimizers import Nadam"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oKC5Rvnt46q2",
        "colab_type": "text"
      },
      "source": [
        "#### Arquitectura\n",
        "Modelo obtenido de https://www.pyimagesearch.com/2020/03/16/detecting-covid-19-in-x-ray-images-with-keras-tensorflow-and-deep-learning/ ."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "udB3lvn7ctL5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Instanciamos el modelo pre-entrenado con los pesos de imagenet\n",
        "baseModel = VGG16(weights=\"imagenet\", include_top=False, \n",
        "                  input_tensor=Input(shape=(224, 224, 3)))\n",
        "\n",
        "# Definimos la parte no convolucional de la red\n",
        "headModel = baseModel.output\n",
        "headModel = AveragePooling2D(pool_size=(4, 4))(headModel)\n",
        "headModel = Flatten(name=\"flatten\")(headModel)\n",
        "headModel = Dense(64, activation=\"relu\")(headModel)\n",
        "headModel = Dropout(0.5)(headModel)\n",
        "headModel = Dense(2, activation=\"softmax\")(headModel)\n",
        "\n",
        "# Componemos el modelo con la parte convolucional pre-entrenada,\n",
        "# junto a la nueva parte de la red densa definida\n",
        "model = Model(inputs=baseModel.input, outputs=headModel)\n",
        "\n",
        "# Congelamos las capas convolucionales para evitar que sean\n",
        "# entrenadas, y poder usar así los filtros de características\n",
        "# ya aprendidos.\n",
        "for layer in baseModel.layers:\n",
        "  layer.trainable = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R1Iwuw7rb1ri",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 969
        },
        "outputId": "4d5e2381-08a8-40b0-a6f6-364a32dbeec4"
      },
      "source": [
        "# https://stackoverflow.com/questions/41665799/keras-model-summary-object-to-string\n",
        "stringlist = []\n",
        "model.summary(print_fn=lambda x: stringlist.append(x))\n",
        "short_model_summary = \"\\n\".join(stringlist)\n",
        "print(short_model_summary)"
      ],
      "execution_count": 63,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_3\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_4 (InputLayer)         [(None, 224, 224, 3)]     0         \n",
            "_________________________________________________________________\n",
            "block1_conv1 (Conv2D)        (None, 224, 224, 64)      1792      \n",
            "_________________________________________________________________\n",
            "block1_conv2 (Conv2D)        (None, 224, 224, 64)      36928     \n",
            "_________________________________________________________________\n",
            "block1_pool (MaxPooling2D)   (None, 112, 112, 64)      0         \n",
            "_________________________________________________________________\n",
            "block2_conv1 (Conv2D)        (None, 112, 112, 128)     73856     \n",
            "_________________________________________________________________\n",
            "block2_conv2 (Conv2D)        (None, 112, 112, 128)     147584    \n",
            "_________________________________________________________________\n",
            "block2_pool (MaxPooling2D)   (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "block3_conv1 (Conv2D)        (None, 56, 56, 256)       295168    \n",
            "_________________________________________________________________\n",
            "block3_conv2 (Conv2D)        (None, 56, 56, 256)       590080    \n",
            "_________________________________________________________________\n",
            "block3_conv3 (Conv2D)        (None, 56, 56, 256)       590080    \n",
            "_________________________________________________________________\n",
            "block3_pool (MaxPooling2D)   (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "block4_conv1 (Conv2D)        (None, 28, 28, 512)       1180160   \n",
            "_________________________________________________________________\n",
            "block4_conv2 (Conv2D)        (None, 28, 28, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block4_conv3 (Conv2D)        (None, 28, 28, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block4_pool (MaxPooling2D)   (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "block5_conv1 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block5_conv2 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block5_conv3 (Conv2D)        (None, 14, 14, 512)       2359808   \n",
            "_________________________________________________________________\n",
            "block5_pool (MaxPooling2D)   (None, 7, 7, 512)         0         \n",
            "_________________________________________________________________\n",
            "average_pooling2d_8 (Average (None, 1, 1, 512)         0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_10 (Dense)             (None, 64)                32832     \n",
            "_________________________________________________________________\n",
            "dropout_7 (Dropout)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_11 (Dense)             (None, 2)                 130       \n",
            "=================================================================\n",
            "Total params: 14,747,650\n",
            "Trainable params: 32,962\n",
            "Non-trainable params: 14,714,688\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7K2cY9h05Amo",
        "colab_type": "text"
      },
      "source": [
        "#### Compilación del modelo"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aj-OQFNoanfU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(loss=\"categorical_crossentropy\", \n",
        "              optimizer=Nadam(),\n",
        "              metrics=[\"accuracy\"])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aX8Xd9Ds5DMn",
        "colab_type": "text"
      },
      "source": [
        "#### Entrenamiento del modelo"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BW5iP3jiWuQY",
        "colab_type": "code",
        "outputId": "05d8a074-3b25-4fac-aba0-21de1c9cb6df",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "H2 = model.fit(train_augmented_data, \n",
        "                        steps_per_epoch=8,\n",
        "                        validation_data=validation_data,\n",
        "                        validation_steps=8,\n",
        "                        epochs=50)"
      ],
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "8/8 [==============================] - 4s 500ms/step - loss: 0.7782 - accuracy: 0.5042 - val_loss: 0.6255 - val_accuracy: 0.7250\n",
            "Epoch 2/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.6879 - accuracy: 0.5975 - val_loss: 0.5506 - val_accuracy: 0.8375\n",
            "Epoch 3/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.5701 - accuracy: 0.7246 - val_loss: 0.4813 - val_accuracy: 0.9375\n",
            "Epoch 4/50\n",
            "8/8 [==============================] - 4s 438ms/step - loss: 0.5578 - accuracy: 0.7203 - val_loss: 0.4330 - val_accuracy: 0.9000\n",
            "Epoch 5/50\n",
            "8/8 [==============================] - 3s 434ms/step - loss: 0.5130 - accuracy: 0.8008 - val_loss: 0.3927 - val_accuracy: 0.8875\n",
            "Epoch 6/50\n",
            "8/8 [==============================] - 3s 434ms/step - loss: 0.4885 - accuracy: 0.7754 - val_loss: 0.3493 - val_accuracy: 0.9125\n",
            "Epoch 7/50\n",
            "8/8 [==============================] - 4s 475ms/step - loss: 0.4397 - accuracy: 0.8305 - val_loss: 0.3213 - val_accuracy: 0.9250\n",
            "Epoch 8/50\n",
            "8/8 [==============================] - 3s 434ms/step - loss: 0.4383 - accuracy: 0.8390 - val_loss: 0.2961 - val_accuracy: 0.9125\n",
            "Epoch 9/50\n",
            "8/8 [==============================] - 4s 441ms/step - loss: 0.3788 - accuracy: 0.8305 - val_loss: 0.2746 - val_accuracy: 0.9125\n",
            "Epoch 10/50\n",
            "8/8 [==============================] - 4s 438ms/step - loss: 0.4415 - accuracy: 0.7797 - val_loss: 0.2587 - val_accuracy: 0.9500\n",
            "Epoch 11/50\n",
            "8/8 [==============================] - 3s 431ms/step - loss: 0.4098 - accuracy: 0.8263 - val_loss: 0.2496 - val_accuracy: 0.9375\n",
            "Epoch 12/50\n",
            "8/8 [==============================] - 3s 435ms/step - loss: 0.3876 - accuracy: 0.8559 - val_loss: 0.2346 - val_accuracy: 0.9375\n",
            "Epoch 13/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3587 - accuracy: 0.8475 - val_loss: 0.2199 - val_accuracy: 0.9625\n",
            "Epoch 14/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3858 - accuracy: 0.8093 - val_loss: 0.2178 - val_accuracy: 0.9500\n",
            "Epoch 15/50\n",
            "8/8 [==============================] - 3s 435ms/step - loss: 0.3042 - accuracy: 0.9068 - val_loss: 0.1981 - val_accuracy: 0.9625\n",
            "Epoch 16/50\n",
            "8/8 [==============================] - 4s 441ms/step - loss: 0.3413 - accuracy: 0.8517 - val_loss: 0.1952 - val_accuracy: 0.9500\n",
            "Epoch 17/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.3246 - accuracy: 0.8475 - val_loss: 0.1810 - val_accuracy: 0.9625\n",
            "Epoch 18/50\n",
            "8/8 [==============================] - 4s 475ms/step - loss: 0.3184 - accuracy: 0.8898 - val_loss: 0.1771 - val_accuracy: 0.9625\n",
            "Epoch 19/50\n",
            "8/8 [==============================] - 3s 430ms/step - loss: 0.3086 - accuracy: 0.8644 - val_loss: 0.1793 - val_accuracy: 0.9625\n",
            "Epoch 20/50\n",
            "8/8 [==============================] - 4s 481ms/step - loss: 0.2830 - accuracy: 0.8814 - val_loss: 0.1683 - val_accuracy: 0.9750\n",
            "Epoch 21/50\n",
            "8/8 [==============================] - 4s 477ms/step - loss: 0.2805 - accuracy: 0.8941 - val_loss: 0.1596 - val_accuracy: 0.9625\n",
            "Epoch 22/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.2616 - accuracy: 0.9025 - val_loss: 0.1532 - val_accuracy: 0.9750\n",
            "Epoch 23/50\n",
            "8/8 [==============================] - 4s 438ms/step - loss: 0.2966 - accuracy: 0.8686 - val_loss: 0.1466 - val_accuracy: 0.9750\n",
            "Epoch 24/50\n",
            "8/8 [==============================] - 3s 431ms/step - loss: 0.2869 - accuracy: 0.8771 - val_loss: 0.1430 - val_accuracy: 0.9750\n",
            "Epoch 25/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.3035 - accuracy: 0.8941 - val_loss: 0.1405 - val_accuracy: 0.9750\n",
            "Epoch 26/50\n",
            "8/8 [==============================] - 3s 432ms/step - loss: 0.2712 - accuracy: 0.8983 - val_loss: 0.1365 - val_accuracy: 0.9750\n",
            "Epoch 27/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.2298 - accuracy: 0.9322 - val_loss: 0.1278 - val_accuracy: 0.9750\n",
            "Epoch 28/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.2296 - accuracy: 0.9237 - val_loss: 0.1426 - val_accuracy: 0.9625\n",
            "Epoch 29/50\n",
            "8/8 [==============================] - 4s 442ms/step - loss: 0.2529 - accuracy: 0.9025 - val_loss: 0.1244 - val_accuracy: 0.9750\n",
            "Epoch 30/50\n",
            "8/8 [==============================] - 4s 446ms/step - loss: 0.2203 - accuracy: 0.9280 - val_loss: 0.1223 - val_accuracy: 0.9750\n",
            "Epoch 31/50\n",
            "8/8 [==============================] - 3s 432ms/step - loss: 0.2349 - accuracy: 0.9110 - val_loss: 0.1204 - val_accuracy: 0.9750\n",
            "Epoch 32/50\n",
            "8/8 [==============================] - 3s 435ms/step - loss: 0.1864 - accuracy: 0.9534 - val_loss: 0.1160 - val_accuracy: 0.9750\n",
            "Epoch 33/50\n",
            "8/8 [==============================] - 3s 434ms/step - loss: 0.1877 - accuracy: 0.9407 - val_loss: 0.1113 - val_accuracy: 0.9625\n",
            "Epoch 34/50\n",
            "8/8 [==============================] - 3s 432ms/step - loss: 0.2043 - accuracy: 0.9237 - val_loss: 0.1105 - val_accuracy: 0.9750\n",
            "Epoch 35/50\n",
            "8/8 [==============================] - 3s 437ms/step - loss: 0.1851 - accuracy: 0.9449 - val_loss: 0.1166 - val_accuracy: 0.9750\n",
            "Epoch 36/50\n",
            "8/8 [==============================] - 4s 444ms/step - loss: 0.1903 - accuracy: 0.9492 - val_loss: 0.1088 - val_accuracy: 0.9750\n",
            "Epoch 37/50\n",
            "8/8 [==============================] - 4s 448ms/step - loss: 0.1730 - accuracy: 0.9407 - val_loss: 0.1056 - val_accuracy: 0.9625\n",
            "Epoch 38/50\n",
            "8/8 [==============================] - 4s 448ms/step - loss: 0.2023 - accuracy: 0.9153 - val_loss: 0.1121 - val_accuracy: 0.9625\n",
            "Epoch 39/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.2086 - accuracy: 0.9110 - val_loss: 0.1014 - val_accuracy: 0.9750\n",
            "Epoch 40/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.1847 - accuracy: 0.9364 - val_loss: 0.1032 - val_accuracy: 0.9750\n",
            "Epoch 41/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.1931 - accuracy: 0.9280 - val_loss: 0.0995 - val_accuracy: 0.9750\n",
            "Epoch 42/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.1625 - accuracy: 0.9492 - val_loss: 0.0985 - val_accuracy: 0.9750\n",
            "Epoch 43/50\n",
            "8/8 [==============================] - 4s 444ms/step - loss: 0.2013 - accuracy: 0.9110 - val_loss: 0.1004 - val_accuracy: 0.9750\n",
            "Epoch 44/50\n",
            "8/8 [==============================] - 4s 438ms/step - loss: 0.1954 - accuracy: 0.9237 - val_loss: 0.1107 - val_accuracy: 0.9625\n",
            "Epoch 45/50\n",
            "8/8 [==============================] - 4s 440ms/step - loss: 0.1506 - accuracy: 0.9492 - val_loss: 0.0957 - val_accuracy: 0.9750\n",
            "Epoch 46/50\n",
            "8/8 [==============================] - 3s 436ms/step - loss: 0.1760 - accuracy: 0.9364 - val_loss: 0.0981 - val_accuracy: 0.9750\n",
            "Epoch 47/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.1577 - accuracy: 0.9661 - val_loss: 0.0955 - val_accuracy: 0.9750\n",
            "Epoch 48/50\n",
            "8/8 [==============================] - 4s 439ms/step - loss: 0.1989 - accuracy: 0.9195 - val_loss: 0.1049 - val_accuracy: 0.9750\n",
            "Epoch 49/50\n",
            "8/8 [==============================] - 3s 435ms/step - loss: 0.1597 - accuracy: 0.9322 - val_loss: 0.0889 - val_accuracy: 0.9750\n",
            "Epoch 50/50\n",
            "8/8 [==============================] - 3s 434ms/step - loss: 0.1536 - accuracy: 0.9492 - val_loss: 0.0891 - val_accuracy: 0.9625\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jzQXhvoh5Gn1",
        "colab_type": "text"
      },
      "source": [
        "#### Visualización de resultados"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RjMO3D4bqfVe",
        "colab_type": "code",
        "outputId": "08b1a54e-e140-4411-ee64-8e12d070ce40",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        }
      },
      "source": [
        "plt.plot(H2.history['accuracy'], label='accuracy')\n",
        "plt.plot(H2.history['val_accuracy'], label = 'val_accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='lower right')"
      ],
      "execution_count": 66,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f98e667ee80>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 66
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S6Dv1jmeqmLV",
        "colab_type": "code",
        "outputId": "faa5dee9-2fd1-454a-b2dc-bfa1306e20ea",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        }
      },
      "source": [
        "plt.plot(H2.history['loss'], label='loss')\n",
        "plt.plot(H2.history['val_loss'], label = 'val_loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss')\n",
        "plt.ylim([0., 1])\n",
        "plt.legend(loc='upper right')"
      ],
      "execution_count": 67,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f98e661eda0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 67
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1zcESuWX58YQ",
        "colab_type": "code",
        "outputId": "c65171d4-6de9-4c71-9a37-75d565299dda",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "model.evaluate(test_data)"
      ],
      "execution_count": 68,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "8/8 [==============================] - 1s 67ms/step - loss: 0.0891 - accuracy: 0.9625\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.08911382406949997, 0.9624999761581421]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 68
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Vhk0V-iGCM4h",
        "colab_type": "code",
        "outputId": "75a17944-10a2-47f2-d4c5-2c546eec03a5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "predictions = model.predict(test_data,\n",
        "                     batch_size=10,\n",
        "                     verbose=1)"
      ],
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "8/8 [==============================] - 1s 66ms/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5SSw-LWTCeLB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_pred = np.argmax(predictions, axis=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CR0Rf6kMCa1B",
        "colab_type": "code",
        "outputId": "50a2953b-2a11-46f6-a164-50a91a48ae99",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "print(confusion_matrix(test_data.classes, y_pred))"
      ],
      "execution_count": 71,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[39  1]\n",
            " [ 2 38]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nP1pH-gVEgrl",
        "colab_type": "code",
        "outputId": "fdf72acf-1dc6-4544-af6e-ba30ef215ad4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 170
        }
      },
      "source": [
        "print(classification_report(test_data.classes, y_pred, target_names = ['covid', 'healthy']))"
      ],
      "execution_count": 72,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "       covid       0.95      0.97      0.96        40\n",
            "     healthy       0.97      0.95      0.96        40\n",
            "\n",
            "    accuracy                           0.96        80\n",
            "   macro avg       0.96      0.96      0.96        80\n",
            "weighted avg       0.96      0.96      0.96        80\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MFlMewdh3xLh",
        "colab_type": "text"
      },
      "source": [
        "### NasNetLarge Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "st41MT014DKA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.applications import NASNetLarge"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aesO_uJZ30rU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "baseModel = NASNetLarge(weights=\"imagenet\", include_top=False, \n",
        "                  input_tensor=Input(shape=(224, 224, 3)))\n",
        "\n",
        "headModel = baseModel.output\n",
        "headModel = AveragePooling2D(pool_size=(4, 4))(headModel)\n",
        "headModel = Flatten(name=\"flatten\")(headModel)\n",
        "headModel = Dense(64, activation=\"relu\")(headModel)\n",
        "headModel = Dropout(0.5)(headModel)\n",
        "headModel = Dense(2, activation=\"softmax\")(headModel)\n",
        "\n",
        "model = Model(inputs=baseModel.input, outputs=headModel)\n",
        "\n",
        "for layer in baseModel.layers:\n",
        "  layer.trainable = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p_JJWaxN4NGE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "b8e8f940-91a6-41a6-dbc6-75bf95487309"
      },
      "source": [
        "stringlist = []\n",
        "model.summary(print_fn=lambda x: stringlist.append(x))\n",
        "short_model_summary = \"\\n\".join(stringlist)\n",
        "print(short_model_summary)"
      ],
      "execution_count": 75,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_4\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_5 (InputLayer)            [(None, 224, 224, 3) 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "stem_conv1 (Conv2D)             (None, 111, 111, 96) 2592        input_5[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "stem_bn1 (BatchNormalization)   (None, 111, 111, 96) 384         stem_conv1[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_463 (Activation)     (None, 111, 111, 96) 0           stem_bn1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "reduction_conv_1_stem_1 (Conv2D (None, 111, 111, 42) 4032        activation_463[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "reduction_bn_1_stem_1 (BatchNor (None, 111, 111, 42) 168         reduction_conv_1_stem_1[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_464 (Activation)     (None, 111, 111, 42) 0           reduction_bn_1_stem_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_466 (Activation)     (None, 111, 111, 96) 0           stem_bn1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 115, 115, 42) 0           activation_464[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 117, 117, 96) 0           activation_466[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 56, 56, 42)   2814        separable_conv_1_pad_reduction_le\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 56, 56, 42)   8736        separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 56, 56, 42)   168         separable_conv_1_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 56, 56, 42)   168         separable_conv_1_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_465 (Activation)     (None, 56, 56, 42)   0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "activation_467 (Activation)     (None, 56, 56, 42)   0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 56, 56, 42)   2814        activation_465[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 56, 56, 42)   3822        activation_467[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_468 (Activation)     (None, 111, 111, 96) 0           stem_bn1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 56, 56, 42)   168         separable_conv_2_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 56, 56, 42)   168         separable_conv_2_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 117, 117, 96) 0           activation_468[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_470 (Activation)     (None, 111, 111, 96) 0           stem_bn1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_1_stem_1 (Add)    (None, 56, 56, 42)   0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 56, 56, 42)   8736        separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 115, 115, 96) 0           activation_470[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_472 (Activation)     (None, 56, 56, 42)   0           reduction_add_1_stem_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 56, 56, 42)   168         separable_conv_1_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 56, 56, 42)   6432        separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 56, 56, 42)   2142        activation_472[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_469 (Activation)     (None, 56, 56, 42)   0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 56, 56, 42)   168         separable_conv_1_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 56, 56, 42)   168         separable_conv_1_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_pad_1_stem_1 (ZeroPad (None, 113, 113, 42) 0           reduction_bn_1_stem_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 56, 56, 42)   3822        activation_469[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_471 (Activation)     (None, 56, 56, 42)   0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "activation_473 (Activation)     (None, 56, 56, 42)   0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left2_stem_1 (MaxPool (None, 56, 56, 42)   0           reduction_pad_1_stem_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 56, 56, 42)   168         separable_conv_2_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 56, 56, 42)   2814        activation_471[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 56, 56, 42)   2142        activation_473[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_relu_1_stem_2 (Activatio (None, 111, 111, 96) 0           stem_bn1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_2_stem_1 (Add)    (None, 56, 56, 42)   0           reduction_left2_stem_1[0][0]     \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left3_stem_1 (Average (None, 56, 56, 42)   0           reduction_pad_1_stem_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 56, 56, 42)   168         separable_conv_2_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left4_stem_1 (Average (None, 56, 56, 42)   0           reduction_add_1_stem_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 56, 56, 42)   168         separable_conv_2_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_right5_stem_1 (MaxPoo (None, 56, 56, 42)   0           reduction_pad_1_stem_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "zero_padding2d_4 (ZeroPadding2D (None, 112, 112, 96) 0           adjust_relu_1_stem_2[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add3_stem_1 (Add)     (None, 56, 56, 42)   0           reduction_left3_stem_1[0][0]     \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "add_4 (Add)                     (None, 56, 56, 42)   0           reduction_add_2_stem_1[0][0]     \n",
            "                                                                 reduction_left4_stem_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add4_stem_1 (Add)     (None, 56, 56, 42)   0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 reduction_right5_stem_1[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "cropping2d_4 (Cropping2D)       (None, 111, 111, 96) 0           zero_padding2d_4[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_concat_stem_1 (Concat (None, 56, 56, 168)  0           reduction_add_2_stem_1[0][0]     \n",
            "                                                                 reduction_add3_stem_1[0][0]      \n",
            "                                                                 add_4[0][0]                      \n",
            "                                                                 reduction_add4_stem_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_1_stem_2 (Avera (None, 56, 56, 96)   0           adjust_relu_1_stem_2[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_2_stem_2 (Avera (None, 56, 56, 96)   0           cropping2d_4[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_474 (Activation)     (None, 56, 56, 168)  0           reduction_concat_stem_1[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_1_stem_2 (Conv2D)   (None, 56, 56, 42)   4032        adjust_avg_pool_1_stem_2[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_2_stem_2 (Conv2D)   (None, 56, 56, 42)   4032        adjust_avg_pool_2_stem_2[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "reduction_conv_1_stem_2 (Conv2D (None, 56, 56, 84)   14112       activation_474[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_4 (Concatenate)     (None, 56, 56, 84)   0           adjust_conv_1_stem_2[0][0]       \n",
            "                                                                 adjust_conv_2_stem_2[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "reduction_bn_1_stem_2 (BatchNor (None, 56, 56, 84)   336         reduction_conv_1_stem_2[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_stem_2 (BatchNormaliz (None, 56, 56, 84)   336         concatenate_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_475 (Activation)     (None, 56, 56, 84)   0           reduction_bn_1_stem_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation_477 (Activation)     (None, 56, 56, 84)   0           adjust_bn_stem_2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 59, 59, 84)   0           activation_475[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 61, 61, 84)   0           activation_477[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 28, 28, 84)   9156        separable_conv_1_pad_reduction_le\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 28, 28, 84)   11172       separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 28, 28, 84)   336         separable_conv_1_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 28, 28, 84)   336         separable_conv_1_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_476 (Activation)     (None, 28, 28, 84)   0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "activation_478 (Activation)     (None, 28, 28, 84)   0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 28, 28, 84)   9156        activation_476[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 28, 28, 84)   11172       activation_478[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_479 (Activation)     (None, 56, 56, 84)   0           adjust_bn_stem_2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 28, 28, 84)   336         separable_conv_2_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 28, 28, 84)   336         separable_conv_2_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 61, 61, 84)   0           activation_479[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_481 (Activation)     (None, 56, 56, 84)   0           adjust_bn_stem_2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_1_stem_2 (Add)    (None, 28, 28, 84)   0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 28, 28, 84)   11172       separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 59, 59, 84)   0           activation_481[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_483 (Activation)     (None, 28, 28, 84)   0           reduction_add_1_stem_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 28, 28, 84)   336         separable_conv_1_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 28, 28, 84)   9156        separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 28, 28, 84)   7812        activation_483[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_480 (Activation)     (None, 28, 28, 84)   0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 28, 28, 84)   336         separable_conv_1_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 28, 28, 84)   336         separable_conv_1_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_pad_1_stem_2 (ZeroPad (None, 57, 57, 84)   0           reduction_bn_1_stem_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 28, 28, 84)   11172       activation_480[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_482 (Activation)     (None, 28, 28, 84)   0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "activation_484 (Activation)     (None, 28, 28, 84)   0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left2_stem_2 (MaxPool (None, 28, 28, 84)   0           reduction_pad_1_stem_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 28, 28, 84)   336         separable_conv_2_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 28, 28, 84)   9156        activation_482[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 28, 28, 84)   7812        activation_484[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_relu_1_0 (Activation)    (None, 56, 56, 168)  0           reduction_concat_stem_1[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_2_stem_2 (Add)    (None, 28, 28, 84)   0           reduction_left2_stem_2[0][0]     \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left3_stem_2 (Average (None, 28, 28, 84)   0           reduction_pad_1_stem_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 28, 28, 84)   336         separable_conv_2_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left4_stem_2 (Average (None, 28, 28, 84)   0           reduction_add_1_stem_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 28, 28, 84)   336         separable_conv_2_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_right5_stem_2 (MaxPoo (None, 28, 28, 84)   0           reduction_pad_1_stem_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "zero_padding2d_5 (ZeroPadding2D (None, 57, 57, 168)  0           adjust_relu_1_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add3_stem_2 (Add)     (None, 28, 28, 84)   0           reduction_left3_stem_2[0][0]     \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "add_5 (Add)                     (None, 28, 28, 84)   0           reduction_add_2_stem_2[0][0]     \n",
            "                                                                 reduction_left4_stem_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add4_stem_2 (Add)     (None, 28, 28, 84)   0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 reduction_right5_stem_2[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "cropping2d_5 (Cropping2D)       (None, 56, 56, 168)  0           zero_padding2d_5[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_concat_stem_2 (Concat (None, 28, 28, 336)  0           reduction_add_2_stem_2[0][0]     \n",
            "                                                                 reduction_add3_stem_2[0][0]      \n",
            "                                                                 add_5[0][0]                      \n",
            "                                                                 reduction_add4_stem_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_1_0 (AveragePoo (None, 28, 28, 168)  0           adjust_relu_1_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_2_0 (AveragePoo (None, 28, 28, 168)  0           cropping2d_5[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_1_0 (Conv2D)        (None, 28, 28, 84)   14112       adjust_avg_pool_1_0[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_2_0 (Conv2D)        (None, 28, 28, 84)   14112       adjust_avg_pool_2_0[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_485 (Activation)     (None, 28, 28, 336)  0           reduction_concat_stem_2[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_5 (Concatenate)     (None, 28, 28, 168)  0           adjust_conv_1_0[0][0]            \n",
            "                                                                 adjust_conv_2_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_0 (Conv2D)        (None, 28, 28, 168)  56448       activation_485[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_0 (BatchNormalization (None, 28, 28, 168)  672         concatenate_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_0 (BatchNormalizati (None, 28, 28, 168)  672         normal_conv_1_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_486 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_0[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_488 (Activation)     (None, 28, 28, 168)  0           adjust_bn_0[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_490 (Activation)     (None, 28, 28, 168)  0           adjust_bn_0[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_492 (Activation)     (None, 28, 28, 168)  0           adjust_bn_0[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_494 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_0[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_0 (None, 28, 28, 168)  32424       activation_486[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 28, 28, 168)  29736       activation_488[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_0 (None, 28, 28, 168)  32424       activation_490[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 28, 28, 168)  29736       activation_492[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_0 (None, 28, 28, 168)  29736       activation_494[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left1_0[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right1_0[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left2_0[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right2_0[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left5_0[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_487 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_489 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_491 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_493 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_495 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_0 (None, 28, 28, 168)  32424       activation_487[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 28, 28, 168)  29736       activation_489[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_0 (None, 28, 28, 168)  32424       activation_491[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 28, 28, 168)  29736       activation_493[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_0 (None, 28, 28, 168)  29736       activation_495[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left1_0[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right1_0[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left2_0[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right2_0[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_0 (AveragePooling2 (None, 28, 28, 168)  0           normal_bn_1_0[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_0 (AveragePooling2 (None, 28, 28, 168)  0           adjust_bn_0[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_0 (AveragePooling (None, 28, 28, 168)  0           adjust_bn_0[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left5_0[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_0 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_0 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_0 (Add)            (None, 28, 28, 168)  0           normal_left3_0[0][0]             \n",
            "                                                                 adjust_bn_0[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_0 (Add)            (None, 28, 28, 168)  0           normal_left4_0[0][0]             \n",
            "                                                                 normal_right4_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_0 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_0[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_0 (Concatenate)   (None, 28, 28, 1008) 0           adjust_bn_0[0][0]                \n",
            "                                                                 normal_add_1_0[0][0]             \n",
            "                                                                 normal_add_2_0[0][0]             \n",
            "                                                                 normal_add_3_0[0][0]             \n",
            "                                                                 normal_add_4_0[0][0]             \n",
            "                                                                 normal_add_5_0[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_496 (Activation)     (None, 28, 28, 336)  0           reduction_concat_stem_2[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_497 (Activation)     (None, 28, 28, 1008) 0           normal_concat_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_1 (Conv2 (None, 28, 28, 168)  56448       activation_496[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_1 (Conv2D)        (None, 28, 28, 168)  169344      activation_497[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_1 (BatchNormalization (None, 28, 28, 168)  672         adjust_conv_projection_1[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_1 (BatchNormalizati (None, 28, 28, 168)  672         normal_conv_1_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_498 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_500 (Activation)     (None, 28, 28, 168)  0           adjust_bn_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_502 (Activation)     (None, 28, 28, 168)  0           adjust_bn_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_504 (Activation)     (None, 28, 28, 168)  0           adjust_bn_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_506 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 28, 28, 168)  32424       activation_498[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 28, 28, 168)  29736       activation_500[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 28, 28, 168)  32424       activation_502[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 28, 28, 168)  29736       activation_504[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 28, 28, 168)  29736       activation_506[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left1_1[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right1_1[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left2_1[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right2_1[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left5_1[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_499 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_501 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_503 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_505 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_507 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 28, 28, 168)  32424       activation_499[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 28, 28, 168)  29736       activation_501[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 28, 28, 168)  32424       activation_503[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 28, 28, 168)  29736       activation_505[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 28, 28, 168)  29736       activation_507[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left1_1[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right1_1[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left2_1[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right2_1[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_1 (AveragePooling2 (None, 28, 28, 168)  0           normal_bn_1_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_1 (AveragePooling2 (None, 28, 28, 168)  0           adjust_bn_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_1 (AveragePooling (None, 28, 28, 168)  0           adjust_bn_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left5_1[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_1 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_1 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_1 (Add)            (None, 28, 28, 168)  0           normal_left3_1[0][0]             \n",
            "                                                                 adjust_bn_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_1 (Add)            (None, 28, 28, 168)  0           normal_left4_1[0][0]             \n",
            "                                                                 normal_right4_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_1 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_1 (Concatenate)   (None, 28, 28, 1008) 0           adjust_bn_1[0][0]                \n",
            "                                                                 normal_add_1_1[0][0]             \n",
            "                                                                 normal_add_2_1[0][0]             \n",
            "                                                                 normal_add_3_1[0][0]             \n",
            "                                                                 normal_add_4_1[0][0]             \n",
            "                                                                 normal_add_5_1[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_508 (Activation)     (None, 28, 28, 1008) 0           normal_concat_0[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_509 (Activation)     (None, 28, 28, 1008) 0           normal_concat_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_2 (Conv2 (None, 28, 28, 168)  169344      activation_508[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_2 (Conv2D)        (None, 28, 28, 168)  169344      activation_509[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_2 (BatchNormalization (None, 28, 28, 168)  672         adjust_conv_projection_2[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_2 (BatchNormalizati (None, 28, 28, 168)  672         normal_conv_1_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_510 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_512 (Activation)     (None, 28, 28, 168)  0           adjust_bn_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_514 (Activation)     (None, 28, 28, 168)  0           adjust_bn_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_516 (Activation)     (None, 28, 28, 168)  0           adjust_bn_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_518 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_2 (None, 28, 28, 168)  32424       activation_510[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 28, 28, 168)  29736       activation_512[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_2 (None, 28, 28, 168)  32424       activation_514[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 28, 28, 168)  29736       activation_516[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_2 (None, 28, 28, 168)  29736       activation_518[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left1_2[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right1_2[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left2_2[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right2_2[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left5_2[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_511 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_513 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_515 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_517 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_519 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_2 (None, 28, 28, 168)  32424       activation_511[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 28, 28, 168)  29736       activation_513[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_2 (None, 28, 28, 168)  32424       activation_515[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 28, 28, 168)  29736       activation_517[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_2 (None, 28, 28, 168)  29736       activation_519[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left1_2[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right1_2[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left2_2[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right2_2[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_2 (AveragePooling2 (None, 28, 28, 168)  0           normal_bn_1_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_2 (AveragePooling2 (None, 28, 28, 168)  0           adjust_bn_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_2 (AveragePooling (None, 28, 28, 168)  0           adjust_bn_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left5_2[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_2 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_2 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_2 (Add)            (None, 28, 28, 168)  0           normal_left3_2[0][0]             \n",
            "                                                                 adjust_bn_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_2 (Add)            (None, 28, 28, 168)  0           normal_left4_2[0][0]             \n",
            "                                                                 normal_right4_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_2 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_2 (Concatenate)   (None, 28, 28, 1008) 0           adjust_bn_2[0][0]                \n",
            "                                                                 normal_add_1_2[0][0]             \n",
            "                                                                 normal_add_2_2[0][0]             \n",
            "                                                                 normal_add_3_2[0][0]             \n",
            "                                                                 normal_add_4_2[0][0]             \n",
            "                                                                 normal_add_5_2[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_520 (Activation)     (None, 28, 28, 1008) 0           normal_concat_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_521 (Activation)     (None, 28, 28, 1008) 0           normal_concat_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_3 (Conv2 (None, 28, 28, 168)  169344      activation_520[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_3 (Conv2D)        (None, 28, 28, 168)  169344      activation_521[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_3 (BatchNormalization (None, 28, 28, 168)  672         adjust_conv_projection_3[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_3 (BatchNormalizati (None, 28, 28, 168)  672         normal_conv_1_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_522 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_524 (Activation)     (None, 28, 28, 168)  0           adjust_bn_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_526 (Activation)     (None, 28, 28, 168)  0           adjust_bn_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_528 (Activation)     (None, 28, 28, 168)  0           adjust_bn_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_530 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_3 (None, 28, 28, 168)  32424       activation_522[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 28, 28, 168)  29736       activation_524[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_3 (None, 28, 28, 168)  32424       activation_526[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 28, 28, 168)  29736       activation_528[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_3 (None, 28, 28, 168)  29736       activation_530[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left1_3[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right1_3[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left2_3[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right2_3[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left5_3[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_523 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_525 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_527 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_529 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_531 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_3 (None, 28, 28, 168)  32424       activation_523[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 28, 28, 168)  29736       activation_525[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_3 (None, 28, 28, 168)  32424       activation_527[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 28, 28, 168)  29736       activation_529[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_3 (None, 28, 28, 168)  29736       activation_531[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left1_3[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right1_3[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left2_3[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right2_3[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_3 (AveragePooling2 (None, 28, 28, 168)  0           normal_bn_1_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_3 (AveragePooling2 (None, 28, 28, 168)  0           adjust_bn_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_3 (AveragePooling (None, 28, 28, 168)  0           adjust_bn_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left5_3[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_3 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_3 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_3 (Add)            (None, 28, 28, 168)  0           normal_left3_3[0][0]             \n",
            "                                                                 adjust_bn_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_3 (Add)            (None, 28, 28, 168)  0           normal_left4_3[0][0]             \n",
            "                                                                 normal_right4_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_3 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_3 (Concatenate)   (None, 28, 28, 1008) 0           adjust_bn_3[0][0]                \n",
            "                                                                 normal_add_1_3[0][0]             \n",
            "                                                                 normal_add_2_3[0][0]             \n",
            "                                                                 normal_add_3_3[0][0]             \n",
            "                                                                 normal_add_4_3[0][0]             \n",
            "                                                                 normal_add_5_3[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_532 (Activation)     (None, 28, 28, 1008) 0           normal_concat_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_533 (Activation)     (None, 28, 28, 1008) 0           normal_concat_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_4 (Conv2 (None, 28, 28, 168)  169344      activation_532[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_4 (Conv2D)        (None, 28, 28, 168)  169344      activation_533[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_4 (BatchNormalization (None, 28, 28, 168)  672         adjust_conv_projection_4[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_4 (BatchNormalizati (None, 28, 28, 168)  672         normal_conv_1_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_534 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_536 (Activation)     (None, 28, 28, 168)  0           adjust_bn_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_538 (Activation)     (None, 28, 28, 168)  0           adjust_bn_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_540 (Activation)     (None, 28, 28, 168)  0           adjust_bn_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_542 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_4 (None, 28, 28, 168)  32424       activation_534[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 28, 28, 168)  29736       activation_536[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_4 (None, 28, 28, 168)  32424       activation_538[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 28, 28, 168)  29736       activation_540[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_4 (None, 28, 28, 168)  29736       activation_542[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left1_4[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right1_4[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left2_4[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right2_4[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left5_4[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_535 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_537 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_539 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_541 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_543 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_4 (None, 28, 28, 168)  32424       activation_535[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 28, 28, 168)  29736       activation_537[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_4 (None, 28, 28, 168)  32424       activation_539[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 28, 28, 168)  29736       activation_541[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_4 (None, 28, 28, 168)  29736       activation_543[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left1_4[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right1_4[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left2_4[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right2_4[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_4 (AveragePooling2 (None, 28, 28, 168)  0           normal_bn_1_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_4 (AveragePooling2 (None, 28, 28, 168)  0           adjust_bn_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_4 (AveragePooling (None, 28, 28, 168)  0           adjust_bn_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left5_4[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_4 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_4 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_4 (Add)            (None, 28, 28, 168)  0           normal_left3_4[0][0]             \n",
            "                                                                 adjust_bn_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_4 (Add)            (None, 28, 28, 168)  0           normal_left4_4[0][0]             \n",
            "                                                                 normal_right4_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_4 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_4 (Concatenate)   (None, 28, 28, 1008) 0           adjust_bn_4[0][0]                \n",
            "                                                                 normal_add_1_4[0][0]             \n",
            "                                                                 normal_add_2_4[0][0]             \n",
            "                                                                 normal_add_3_4[0][0]             \n",
            "                                                                 normal_add_4_4[0][0]             \n",
            "                                                                 normal_add_5_4[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_544 (Activation)     (None, 28, 28, 1008) 0           normal_concat_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_545 (Activation)     (None, 28, 28, 1008) 0           normal_concat_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_5 (Conv2 (None, 28, 28, 168)  169344      activation_544[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_5 (Conv2D)        (None, 28, 28, 168)  169344      activation_545[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_5 (BatchNormalization (None, 28, 28, 168)  672         adjust_conv_projection_5[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_5 (BatchNormalizati (None, 28, 28, 168)  672         normal_conv_1_5[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_546 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_548 (Activation)     (None, 28, 28, 168)  0           adjust_bn_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_550 (Activation)     (None, 28, 28, 168)  0           adjust_bn_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_552 (Activation)     (None, 28, 28, 168)  0           adjust_bn_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_554 (Activation)     (None, 28, 28, 168)  0           normal_bn_1_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_5 (None, 28, 28, 168)  32424       activation_546[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 28, 28, 168)  29736       activation_548[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_5 (None, 28, 28, 168)  32424       activation_550[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 28, 28, 168)  29736       activation_552[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_5 (None, 28, 28, 168)  29736       activation_554[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left1_5[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right1_5[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left2_5[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_1_normal_right2_5[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 28, 28, 168)  672         separable_conv_1_normal_left5_5[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_547 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_549 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_551 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_553 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_555 (Activation)     (None, 28, 28, 168)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_5 (None, 28, 28, 168)  32424       activation_547[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 28, 28, 168)  29736       activation_549[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_5 (None, 28, 28, 168)  32424       activation_551[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 28, 28, 168)  29736       activation_553[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_5 (None, 28, 28, 168)  29736       activation_555[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left1_5[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right1_5[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left2_5[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 28, 28, 168)  672         separable_conv_2_normal_right2_5[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_5 (AveragePooling2 (None, 28, 28, 168)  0           normal_bn_1_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_5 (AveragePooling2 (None, 28, 28, 168)  0           adjust_bn_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_5 (AveragePooling (None, 28, 28, 168)  0           adjust_bn_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 28, 28, 168)  672         separable_conv_2_normal_left5_5[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_5 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_5 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_5 (Add)            (None, 28, 28, 168)  0           normal_left3_5[0][0]             \n",
            "                                                                 adjust_bn_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_5 (Add)            (None, 28, 28, 168)  0           normal_left4_5[0][0]             \n",
            "                                                                 normal_right4_5[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_5 (Add)            (None, 28, 28, 168)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_5 (Concatenate)   (None, 28, 28, 1008) 0           adjust_bn_5[0][0]                \n",
            "                                                                 normal_add_1_5[0][0]             \n",
            "                                                                 normal_add_2_5[0][0]             \n",
            "                                                                 normal_add_3_5[0][0]             \n",
            "                                                                 normal_add_4_5[0][0]             \n",
            "                                                                 normal_add_5_5[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_557 (Activation)     (None, 28, 28, 1008) 0           normal_concat_5[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_556 (Activation)     (None, 28, 28, 1008) 0           normal_concat_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "reduction_conv_1_reduce_6 (Conv (None, 28, 28, 336)  338688      activation_557[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_reduce_6 (None, 28, 28, 336)  338688      activation_556[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "reduction_bn_1_reduce_6 (BatchN (None, 28, 28, 336)  1344        reduction_conv_1_reduce_6[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_reduce_6 (BatchNormal (None, 28, 28, 336)  1344        adjust_conv_projection_reduce_6[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_558 (Activation)     (None, 28, 28, 336)  0           reduction_bn_1_reduce_6[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_560 (Activation)     (None, 28, 28, 336)  0           adjust_bn_reduce_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 31, 31, 336)  0           activation_558[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 33, 33, 336)  0           activation_560[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 14, 14, 336)  121296      separable_conv_1_pad_reduction_le\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 14, 14, 336)  129360      separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 14, 14, 336)  1344        separable_conv_1_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 14, 14, 336)  1344        separable_conv_1_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_559 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "activation_561 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 14, 14, 336)  121296      activation_559[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 14, 14, 336)  129360      activation_561[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_562 (Activation)     (None, 28, 28, 336)  0           adjust_bn_reduce_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 14, 14, 336)  1344        separable_conv_2_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 14, 14, 336)  1344        separable_conv_2_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 33, 33, 336)  0           activation_562[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_564 (Activation)     (None, 28, 28, 336)  0           adjust_bn_reduce_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_1_reduce_6 (Add)  (None, 14, 14, 336)  0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 14, 14, 336)  129360      separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 31, 31, 336)  0           activation_564[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_566 (Activation)     (None, 14, 14, 336)  0           reduction_add_1_reduce_6[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 14, 14, 336)  1344        separable_conv_1_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 14, 14, 336)  121296      separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 14, 14, 336)  115920      activation_566[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_563 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 14, 14, 336)  1344        separable_conv_1_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 14, 14, 336)  1344        separable_conv_1_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_pad_1_reduce_6 (ZeroP (None, 29, 29, 336)  0           reduction_bn_1_reduce_6[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 14, 14, 336)  129360      activation_563[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_565 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "activation_567 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left2_reduce_6 (MaxPo (None, 14, 14, 336)  0           reduction_pad_1_reduce_6[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 14, 14, 336)  1344        separable_conv_2_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 14, 14, 336)  121296      activation_565[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 14, 14, 336)  115920      activation_567[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_relu_1_7 (Activation)    (None, 28, 28, 1008) 0           normal_concat_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_2_reduce_6 (Add)  (None, 14, 14, 336)  0           reduction_left2_reduce_6[0][0]   \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left3_reduce_6 (Avera (None, 14, 14, 336)  0           reduction_pad_1_reduce_6[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 14, 14, 336)  1344        separable_conv_2_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left4_reduce_6 (Avera (None, 14, 14, 336)  0           reduction_add_1_reduce_6[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 14, 14, 336)  1344        separable_conv_2_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_right5_reduce_6 (MaxP (None, 14, 14, 336)  0           reduction_pad_1_reduce_6[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "zero_padding2d_6 (ZeroPadding2D (None, 29, 29, 1008) 0           adjust_relu_1_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add3_reduce_6 (Add)   (None, 14, 14, 336)  0           reduction_left3_reduce_6[0][0]   \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "add_6 (Add)                     (None, 14, 14, 336)  0           reduction_add_2_reduce_6[0][0]   \n",
            "                                                                 reduction_left4_reduce_6[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add4_reduce_6 (Add)   (None, 14, 14, 336)  0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 reduction_right5_reduce_6[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "cropping2d_6 (Cropping2D)       (None, 28, 28, 1008) 0           zero_padding2d_6[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_concat_reduce_6 (Conc (None, 14, 14, 1344) 0           reduction_add_2_reduce_6[0][0]   \n",
            "                                                                 reduction_add3_reduce_6[0][0]    \n",
            "                                                                 add_6[0][0]                      \n",
            "                                                                 reduction_add4_reduce_6[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_1_7 (AveragePoo (None, 14, 14, 1008) 0           adjust_relu_1_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_2_7 (AveragePoo (None, 14, 14, 1008) 0           cropping2d_6[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_1_7 (Conv2D)        (None, 14, 14, 168)  169344      adjust_avg_pool_1_7[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_2_7 (Conv2D)        (None, 14, 14, 168)  169344      adjust_avg_pool_2_7[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "activation_568 (Activation)     (None, 14, 14, 1344) 0           reduction_concat_reduce_6[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_6 (Concatenate)     (None, 14, 14, 336)  0           adjust_conv_1_7[0][0]            \n",
            "                                                                 adjust_conv_2_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_7 (Conv2D)        (None, 14, 14, 336)  451584      activation_568[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_7 (BatchNormalization (None, 14, 14, 336)  1344        concatenate_6[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_7 (BatchNormalizati (None, 14, 14, 336)  1344        normal_conv_1_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_569 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_571 (Activation)     (None, 14, 14, 336)  0           adjust_bn_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_573 (Activation)     (None, 14, 14, 336)  0           adjust_bn_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_575 (Activation)     (None, 14, 14, 336)  0           adjust_bn_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_577 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_7 (None, 14, 14, 336)  121296      activation_569[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 14, 14, 336)  115920      activation_571[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_7 (None, 14, 14, 336)  121296      activation_573[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 14, 14, 336)  115920      activation_575[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_7 (None, 14, 14, 336)  115920      activation_577[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left1_7[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right1_7[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left2_7[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right2_7[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left5_7[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_570 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_572 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_574 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_576 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_578 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_7 (None, 14, 14, 336)  121296      activation_570[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 14, 14, 336)  115920      activation_572[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_7 (None, 14, 14, 336)  121296      activation_574[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 14, 14, 336)  115920      activation_576[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_7 (None, 14, 14, 336)  115920      activation_578[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left1_7[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right1_7[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left2_7[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right2_7[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_7 (AveragePooling2 (None, 14, 14, 336)  0           normal_bn_1_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_7 (AveragePooling2 (None, 14, 14, 336)  0           adjust_bn_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_7 (AveragePooling (None, 14, 14, 336)  0           adjust_bn_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left5_7[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_7 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_7 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_7 (Add)            (None, 14, 14, 336)  0           normal_left3_7[0][0]             \n",
            "                                                                 adjust_bn_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_7 (Add)            (None, 14, 14, 336)  0           normal_left4_7[0][0]             \n",
            "                                                                 normal_right4_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_7 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_7 (Concatenate)   (None, 14, 14, 2016) 0           adjust_bn_7[0][0]                \n",
            "                                                                 normal_add_1_7[0][0]             \n",
            "                                                                 normal_add_2_7[0][0]             \n",
            "                                                                 normal_add_3_7[0][0]             \n",
            "                                                                 normal_add_4_7[0][0]             \n",
            "                                                                 normal_add_5_7[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_579 (Activation)     (None, 14, 14, 1344) 0           reduction_concat_reduce_6[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "activation_580 (Activation)     (None, 14, 14, 2016) 0           normal_concat_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_8 (Conv2 (None, 14, 14, 336)  451584      activation_579[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_8 (Conv2D)        (None, 14, 14, 336)  677376      activation_580[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_8 (BatchNormalization (None, 14, 14, 336)  1344        adjust_conv_projection_8[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_8 (BatchNormalizati (None, 14, 14, 336)  1344        normal_conv_1_8[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_581 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_8[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_583 (Activation)     (None, 14, 14, 336)  0           adjust_bn_8[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_585 (Activation)     (None, 14, 14, 336)  0           adjust_bn_8[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_587 (Activation)     (None, 14, 14, 336)  0           adjust_bn_8[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_589 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_8[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_8 (None, 14, 14, 336)  121296      activation_581[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 14, 14, 336)  115920      activation_583[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_8 (None, 14, 14, 336)  121296      activation_585[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 14, 14, 336)  115920      activation_587[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_8 (None, 14, 14, 336)  115920      activation_589[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left1_8[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right1_8[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left2_8[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right2_8[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left5_8[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_582 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_584 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_586 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_588 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_590 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_8 (None, 14, 14, 336)  121296      activation_582[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 14, 14, 336)  115920      activation_584[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_8 (None, 14, 14, 336)  121296      activation_586[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 14, 14, 336)  115920      activation_588[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_8 (None, 14, 14, 336)  115920      activation_590[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left1_8[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right1_8[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left2_8[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right2_8[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_8 (AveragePooling2 (None, 14, 14, 336)  0           normal_bn_1_8[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_8 (AveragePooling2 (None, 14, 14, 336)  0           adjust_bn_8[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_8 (AveragePooling (None, 14, 14, 336)  0           adjust_bn_8[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left5_8[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_8 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_8 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_8 (Add)            (None, 14, 14, 336)  0           normal_left3_8[0][0]             \n",
            "                                                                 adjust_bn_8[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_8 (Add)            (None, 14, 14, 336)  0           normal_left4_8[0][0]             \n",
            "                                                                 normal_right4_8[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_8 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_8[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_8 (Concatenate)   (None, 14, 14, 2016) 0           adjust_bn_8[0][0]                \n",
            "                                                                 normal_add_1_8[0][0]             \n",
            "                                                                 normal_add_2_8[0][0]             \n",
            "                                                                 normal_add_3_8[0][0]             \n",
            "                                                                 normal_add_4_8[0][0]             \n",
            "                                                                 normal_add_5_8[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_591 (Activation)     (None, 14, 14, 2016) 0           normal_concat_7[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_592 (Activation)     (None, 14, 14, 2016) 0           normal_concat_8[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_9 (Conv2 (None, 14, 14, 336)  677376      activation_591[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_9 (Conv2D)        (None, 14, 14, 336)  677376      activation_592[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_9 (BatchNormalization (None, 14, 14, 336)  1344        adjust_conv_projection_9[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_9 (BatchNormalizati (None, 14, 14, 336)  1344        normal_conv_1_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_593 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_9[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_595 (Activation)     (None, 14, 14, 336)  0           adjust_bn_9[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_597 (Activation)     (None, 14, 14, 336)  0           adjust_bn_9[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_599 (Activation)     (None, 14, 14, 336)  0           adjust_bn_9[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "activation_601 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_9[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_9 (None, 14, 14, 336)  121296      activation_593[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 14, 14, 336)  115920      activation_595[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_9 (None, 14, 14, 336)  121296      activation_597[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 14, 14, 336)  115920      activation_599[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_9 (None, 14, 14, 336)  115920      activation_601[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left1_9[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right1_9[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left2_9[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right2_9[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left5_9[0\n",
            "__________________________________________________________________________________________________\n",
            "activation_594 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_596 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_598 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_600 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_602 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_9 (None, 14, 14, 336)  121296      activation_594[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 14, 14, 336)  115920      activation_596[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_9 (None, 14, 14, 336)  121296      activation_598[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 14, 14, 336)  115920      activation_600[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_9 (None, 14, 14, 336)  115920      activation_602[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left1_9[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right1_9[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left2_9[0\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right2_9[\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_9 (AveragePooling2 (None, 14, 14, 336)  0           normal_bn_1_9[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_9 (AveragePooling2 (None, 14, 14, 336)  0           adjust_bn_9[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_9 (AveragePooling (None, 14, 14, 336)  0           adjust_bn_9[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left5_9[0\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_9 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_9 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_9 (Add)            (None, 14, 14, 336)  0           normal_left3_9[0][0]             \n",
            "                                                                 adjust_bn_9[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_9 (Add)            (None, 14, 14, 336)  0           normal_left4_9[0][0]             \n",
            "                                                                 normal_right4_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_9 (Add)            (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_9[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_9 (Concatenate)   (None, 14, 14, 2016) 0           adjust_bn_9[0][0]                \n",
            "                                                                 normal_add_1_9[0][0]             \n",
            "                                                                 normal_add_2_9[0][0]             \n",
            "                                                                 normal_add_3_9[0][0]             \n",
            "                                                                 normal_add_4_9[0][0]             \n",
            "                                                                 normal_add_5_9[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_603 (Activation)     (None, 14, 14, 2016) 0           normal_concat_8[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_604 (Activation)     (None, 14, 14, 2016) 0           normal_concat_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_10 (Conv (None, 14, 14, 336)  677376      activation_603[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_10 (Conv2D)       (None, 14, 14, 336)  677376      activation_604[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_10 (BatchNormalizatio (None, 14, 14, 336)  1344        adjust_conv_projection_10[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_10 (BatchNormalizat (None, 14, 14, 336)  1344        normal_conv_1_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_605 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_10[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_607 (Activation)     (None, 14, 14, 336)  0           adjust_bn_10[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_609 (Activation)     (None, 14, 14, 336)  0           adjust_bn_10[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_611 (Activation)     (None, 14, 14, 336)  0           adjust_bn_10[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_613 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_10[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 14, 14, 336)  121296      activation_605[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 14, 14, 336)  115920      activation_607[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 14, 14, 336)  121296      activation_609[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 14, 14, 336)  115920      activation_611[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 14, 14, 336)  115920      activation_613[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left1_10[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right1_10\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left2_10[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right2_10\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left5_10[\n",
            "__________________________________________________________________________________________________\n",
            "activation_606 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_608 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_610 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_612 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_614 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 14, 14, 336)  121296      activation_606[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 14, 14, 336)  115920      activation_608[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 14, 14, 336)  121296      activation_610[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 14, 14, 336)  115920      activation_612[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 14, 14, 336)  115920      activation_614[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left1_10[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right1_10\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left2_10[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right2_10\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_10 (AveragePooling (None, 14, 14, 336)  0           normal_bn_1_10[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_10 (AveragePooling (None, 14, 14, 336)  0           adjust_bn_10[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_10 (AveragePoolin (None, 14, 14, 336)  0           adjust_bn_10[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left5_10[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_10 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_10 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_10 (Add)           (None, 14, 14, 336)  0           normal_left3_10[0][0]            \n",
            "                                                                 adjust_bn_10[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_10 (Add)           (None, 14, 14, 336)  0           normal_left4_10[0][0]            \n",
            "                                                                 normal_right4_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_10 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_10[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_10 (Concatenate)  (None, 14, 14, 2016) 0           adjust_bn_10[0][0]               \n",
            "                                                                 normal_add_1_10[0][0]            \n",
            "                                                                 normal_add_2_10[0][0]            \n",
            "                                                                 normal_add_3_10[0][0]            \n",
            "                                                                 normal_add_4_10[0][0]            \n",
            "                                                                 normal_add_5_10[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_615 (Activation)     (None, 14, 14, 2016) 0           normal_concat_9[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_616 (Activation)     (None, 14, 14, 2016) 0           normal_concat_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_11 (Conv (None, 14, 14, 336)  677376      activation_615[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_11 (Conv2D)       (None, 14, 14, 336)  677376      activation_616[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_11 (BatchNormalizatio (None, 14, 14, 336)  1344        adjust_conv_projection_11[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_11 (BatchNormalizat (None, 14, 14, 336)  1344        normal_conv_1_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_617 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_11[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_619 (Activation)     (None, 14, 14, 336)  0           adjust_bn_11[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_621 (Activation)     (None, 14, 14, 336)  0           adjust_bn_11[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_623 (Activation)     (None, 14, 14, 336)  0           adjust_bn_11[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_625 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_11[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 14, 14, 336)  121296      activation_617[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 14, 14, 336)  115920      activation_619[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 14, 14, 336)  121296      activation_621[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 14, 14, 336)  115920      activation_623[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 14, 14, 336)  115920      activation_625[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left1_11[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right1_11\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left2_11[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right2_11\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left5_11[\n",
            "__________________________________________________________________________________________________\n",
            "activation_618 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_620 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_622 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_624 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_626 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 14, 14, 336)  121296      activation_618[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 14, 14, 336)  115920      activation_620[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 14, 14, 336)  121296      activation_622[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 14, 14, 336)  115920      activation_624[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 14, 14, 336)  115920      activation_626[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left1_11[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right1_11\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left2_11[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right2_11\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_11 (AveragePooling (None, 14, 14, 336)  0           normal_bn_1_11[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_11 (AveragePooling (None, 14, 14, 336)  0           adjust_bn_11[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_11 (AveragePoolin (None, 14, 14, 336)  0           adjust_bn_11[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left5_11[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_11 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_11 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_11 (Add)           (None, 14, 14, 336)  0           normal_left3_11[0][0]            \n",
            "                                                                 adjust_bn_11[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_11 (Add)           (None, 14, 14, 336)  0           normal_left4_11[0][0]            \n",
            "                                                                 normal_right4_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_11 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_11[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_11 (Concatenate)  (None, 14, 14, 2016) 0           adjust_bn_11[0][0]               \n",
            "                                                                 normal_add_1_11[0][0]            \n",
            "                                                                 normal_add_2_11[0][0]            \n",
            "                                                                 normal_add_3_11[0][0]            \n",
            "                                                                 normal_add_4_11[0][0]            \n",
            "                                                                 normal_add_5_11[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_627 (Activation)     (None, 14, 14, 2016) 0           normal_concat_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_628 (Activation)     (None, 14, 14, 2016) 0           normal_concat_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_12 (Conv (None, 14, 14, 336)  677376      activation_627[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_12 (Conv2D)       (None, 14, 14, 336)  677376      activation_628[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_12 (BatchNormalizatio (None, 14, 14, 336)  1344        adjust_conv_projection_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_12 (BatchNormalizat (None, 14, 14, 336)  1344        normal_conv_1_12[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_629 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_12[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_631 (Activation)     (None, 14, 14, 336)  0           adjust_bn_12[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_633 (Activation)     (None, 14, 14, 336)  0           adjust_bn_12[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_635 (Activation)     (None, 14, 14, 336)  0           adjust_bn_12[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_637 (Activation)     (None, 14, 14, 336)  0           normal_bn_1_12[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 14, 14, 336)  121296      activation_629[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 14, 14, 336)  115920      activation_631[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 14, 14, 336)  121296      activation_633[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 14, 14, 336)  115920      activation_635[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 14, 14, 336)  115920      activation_637[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left1_12[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right1_12\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left2_12[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_1_normal_right2_12\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_1_normal_left5_12[\n",
            "__________________________________________________________________________________________________\n",
            "activation_630 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_632 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_634 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_636 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_638 (Activation)     (None, 14, 14, 336)  0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 14, 14, 336)  121296      activation_630[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 14, 14, 336)  115920      activation_632[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 14, 14, 336)  121296      activation_634[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 14, 14, 336)  115920      activation_636[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 14, 14, 336)  115920      activation_638[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left1_12[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right1_12\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left2_12[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 14, 14, 336)  1344        separable_conv_2_normal_right2_12\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_12 (AveragePooling (None, 14, 14, 336)  0           normal_bn_1_12[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_12 (AveragePooling (None, 14, 14, 336)  0           adjust_bn_12[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_12 (AveragePoolin (None, 14, 14, 336)  0           adjust_bn_12[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 14, 14, 336)  1344        separable_conv_2_normal_left5_12[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_12 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_12 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_12 (Add)           (None, 14, 14, 336)  0           normal_left3_12[0][0]            \n",
            "                                                                 adjust_bn_12[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_12 (Add)           (None, 14, 14, 336)  0           normal_left4_12[0][0]            \n",
            "                                                                 normal_right4_12[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_12 (Add)           (None, 14, 14, 336)  0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_12[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_12 (Concatenate)  (None, 14, 14, 2016) 0           adjust_bn_12[0][0]               \n",
            "                                                                 normal_add_1_12[0][0]            \n",
            "                                                                 normal_add_2_12[0][0]            \n",
            "                                                                 normal_add_3_12[0][0]            \n",
            "                                                                 normal_add_4_12[0][0]            \n",
            "                                                                 normal_add_5_12[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_640 (Activation)     (None, 14, 14, 2016) 0           normal_concat_12[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_639 (Activation)     (None, 14, 14, 2016) 0           normal_concat_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_conv_1_reduce_12 (Con (None, 14, 14, 672)  1354752     activation_640[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_reduce_1 (None, 14, 14, 672)  1354752     activation_639[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "reduction_bn_1_reduce_12 (Batch (None, 14, 14, 672)  2688        reduction_conv_1_reduce_12[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_reduce_12 (BatchNorma (None, 14, 14, 672)  2688        adjust_conv_projection_reduce_12[\n",
            "__________________________________________________________________________________________________\n",
            "activation_641 (Activation)     (None, 14, 14, 672)  0           reduction_bn_1_reduce_12[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "activation_643 (Activation)     (None, 14, 14, 672)  0           adjust_bn_reduce_12[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 17, 17, 672)  0           activation_641[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 19, 19, 672)  0           activation_643[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 7, 7, 672)    468384      separable_conv_1_pad_reduction_le\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 7, 7, 672)    484512      separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 7, 7, 672)    2688        separable_conv_1_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 7, 7, 672)    2688        separable_conv_1_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_642 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "activation_644 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 7, 7, 672)    468384      activation_642[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 7, 7, 672)    484512      activation_644[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_645 (Activation)     (None, 14, 14, 672)  0           adjust_bn_reduce_12[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 7, 7, 672)    2688        separable_conv_2_reduction_left1_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 7, 7, 672)    2688        separable_conv_2_reduction_right1\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 19, 19, 672)  0           activation_645[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_647 (Activation)     (None, 14, 14, 672)  0           adjust_bn_reduce_12[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_1_reduce_12 (Add) (None, 7, 7, 672)    0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 7, 7, 672)    484512      separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_pad_reduction_ (None, 17, 17, 672)  0           activation_647[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_649 (Activation)     (None, 7, 7, 672)    0           reduction_add_1_reduce_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 7, 7, 672)    2688        separable_conv_1_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_righ (None, 7, 7, 672)    468384      separable_conv_1_pad_reduction_ri\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_reduction_left (None, 7, 7, 672)    457632      activation_649[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_646 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_r (None, 7, 7, 672)    2688        separable_conv_1_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_reduction_l (None, 7, 7, 672)    2688        separable_conv_1_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_pad_1_reduce_12 (Zero (None, 15, 15, 672)  0           reduction_bn_1_reduce_12[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 7, 7, 672)    484512      activation_646[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_648 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "activation_650 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_reduction_lef\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left2_reduce_12 (MaxP (None, 7, 7, 672)    0           reduction_pad_1_reduce_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 7, 7, 672)    2688        separable_conv_2_reduction_right2\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_righ (None, 7, 7, 672)    468384      activation_648[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_reduction_left (None, 7, 7, 672)    457632      activation_650[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_relu_1_13 (Activation)   (None, 14, 14, 2016) 0           normal_concat_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add_2_reduce_12 (Add) (None, 7, 7, 672)    0           reduction_left2_reduce_12[0][0]  \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left3_reduce_12 (Aver (None, 7, 7, 672)    0           reduction_pad_1_reduce_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_r (None, 7, 7, 672)    2688        separable_conv_2_reduction_right3\n",
            "__________________________________________________________________________________________________\n",
            "reduction_left4_reduce_12 (Aver (None, 7, 7, 672)    0           reduction_add_1_reduce_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_reduction_l (None, 7, 7, 672)    2688        separable_conv_2_reduction_left4_\n",
            "__________________________________________________________________________________________________\n",
            "reduction_right5_reduce_12 (Max (None, 7, 7, 672)    0           reduction_pad_1_reduce_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "zero_padding2d_7 (ZeroPadding2D (None, 15, 15, 2016) 0           adjust_relu_1_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add3_reduce_12 (Add)  (None, 7, 7, 672)    0           reduction_left3_reduce_12[0][0]  \n",
            "                                                                 separable_conv_2_bn_reduction_rig\n",
            "__________________________________________________________________________________________________\n",
            "add_7 (Add)                     (None, 7, 7, 672)    0           reduction_add_2_reduce_12[0][0]  \n",
            "                                                                 reduction_left4_reduce_12[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "reduction_add4_reduce_12 (Add)  (None, 7, 7, 672)    0           separable_conv_2_bn_reduction_lef\n",
            "                                                                 reduction_right5_reduce_12[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "cropping2d_7 (Cropping2D)       (None, 14, 14, 2016) 0           zero_padding2d_7[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "reduction_concat_reduce_12 (Con (None, 7, 7, 2688)   0           reduction_add_2_reduce_12[0][0]  \n",
            "                                                                 reduction_add3_reduce_12[0][0]   \n",
            "                                                                 add_7[0][0]                      \n",
            "                                                                 reduction_add4_reduce_12[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_1_13 (AveragePo (None, 7, 7, 2016)   0           adjust_relu_1_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_avg_pool_2_13 (AveragePo (None, 7, 7, 2016)   0           cropping2d_7[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_1_13 (Conv2D)       (None, 7, 7, 336)    677376      adjust_avg_pool_1_13[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_2_13 (Conv2D)       (None, 7, 7, 336)    677376      adjust_avg_pool_2_13[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "activation_651 (Activation)     (None, 7, 7, 2688)   0           reduction_concat_reduce_12[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_7 (Concatenate)     (None, 7, 7, 672)    0           adjust_conv_1_13[0][0]           \n",
            "                                                                 adjust_conv_2_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_13 (Conv2D)       (None, 7, 7, 672)    1806336     activation_651[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_13 (BatchNormalizatio (None, 7, 7, 672)    2688        concatenate_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_13 (BatchNormalizat (None, 7, 7, 672)    2688        normal_conv_1_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_652 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_13[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_654 (Activation)     (None, 7, 7, 672)    0           adjust_bn_13[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_656 (Activation)     (None, 7, 7, 672)    0           adjust_bn_13[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_658 (Activation)     (None, 7, 7, 672)    0           adjust_bn_13[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_660 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_13[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 7, 7, 672)    468384      activation_652[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 7, 7, 672)    457632      activation_654[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 7, 7, 672)    468384      activation_656[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 7, 7, 672)    457632      activation_658[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 7, 7, 672)    457632      activation_660[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left1_13[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right1_13\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left2_13[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right2_13\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left5_13[\n",
            "__________________________________________________________________________________________________\n",
            "activation_653 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_655 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_657 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_659 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_661 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 7, 7, 672)    468384      activation_653[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 7, 7, 672)    457632      activation_655[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 7, 7, 672)    468384      activation_657[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 7, 7, 672)    457632      activation_659[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 7, 7, 672)    457632      activation_661[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left1_13[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right1_13\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left2_13[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right2_13\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_13 (AveragePooling (None, 7, 7, 672)    0           normal_bn_1_13[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_13 (AveragePooling (None, 7, 7, 672)    0           adjust_bn_13[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_13 (AveragePoolin (None, 7, 7, 672)    0           adjust_bn_13[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left5_13[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_13 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_13 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_13 (Add)           (None, 7, 7, 672)    0           normal_left3_13[0][0]            \n",
            "                                                                 adjust_bn_13[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_13 (Add)           (None, 7, 7, 672)    0           normal_left4_13[0][0]            \n",
            "                                                                 normal_right4_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_13 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_13[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_13 (Concatenate)  (None, 7, 7, 4032)   0           adjust_bn_13[0][0]               \n",
            "                                                                 normal_add_1_13[0][0]            \n",
            "                                                                 normal_add_2_13[0][0]            \n",
            "                                                                 normal_add_3_13[0][0]            \n",
            "                                                                 normal_add_4_13[0][0]            \n",
            "                                                                 normal_add_5_13[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_662 (Activation)     (None, 7, 7, 2688)   0           reduction_concat_reduce_12[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "activation_663 (Activation)     (None, 7, 7, 4032)   0           normal_concat_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_14 (Conv (None, 7, 7, 672)    1806336     activation_662[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_14 (Conv2D)       (None, 7, 7, 672)    2709504     activation_663[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_14 (BatchNormalizatio (None, 7, 7, 672)    2688        adjust_conv_projection_14[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_14 (BatchNormalizat (None, 7, 7, 672)    2688        normal_conv_1_14[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_664 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_14[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_666 (Activation)     (None, 7, 7, 672)    0           adjust_bn_14[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_668 (Activation)     (None, 7, 7, 672)    0           adjust_bn_14[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_670 (Activation)     (None, 7, 7, 672)    0           adjust_bn_14[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_672 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_14[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 7, 7, 672)    468384      activation_664[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 7, 7, 672)    457632      activation_666[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 7, 7, 672)    468384      activation_668[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 7, 7, 672)    457632      activation_670[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 7, 7, 672)    457632      activation_672[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left1_14[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right1_14\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left2_14[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right2_14\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left5_14[\n",
            "__________________________________________________________________________________________________\n",
            "activation_665 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_667 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_669 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_671 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_673 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 7, 7, 672)    468384      activation_665[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 7, 7, 672)    457632      activation_667[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 7, 7, 672)    468384      activation_669[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 7, 7, 672)    457632      activation_671[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 7, 7, 672)    457632      activation_673[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left1_14[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right1_14\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left2_14[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right2_14\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_14 (AveragePooling (None, 7, 7, 672)    0           normal_bn_1_14[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_14 (AveragePooling (None, 7, 7, 672)    0           adjust_bn_14[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_14 (AveragePoolin (None, 7, 7, 672)    0           adjust_bn_14[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left5_14[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_14 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_14 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_14 (Add)           (None, 7, 7, 672)    0           normal_left3_14[0][0]            \n",
            "                                                                 adjust_bn_14[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_14 (Add)           (None, 7, 7, 672)    0           normal_left4_14[0][0]            \n",
            "                                                                 normal_right4_14[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_14 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_14[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_14 (Concatenate)  (None, 7, 7, 4032)   0           adjust_bn_14[0][0]               \n",
            "                                                                 normal_add_1_14[0][0]            \n",
            "                                                                 normal_add_2_14[0][0]            \n",
            "                                                                 normal_add_3_14[0][0]            \n",
            "                                                                 normal_add_4_14[0][0]            \n",
            "                                                                 normal_add_5_14[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_674 (Activation)     (None, 7, 7, 4032)   0           normal_concat_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_675 (Activation)     (None, 7, 7, 4032)   0           normal_concat_14[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_15 (Conv (None, 7, 7, 672)    2709504     activation_674[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_15 (Conv2D)       (None, 7, 7, 672)    2709504     activation_675[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_15 (BatchNormalizatio (None, 7, 7, 672)    2688        adjust_conv_projection_15[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_15 (BatchNormalizat (None, 7, 7, 672)    2688        normal_conv_1_15[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_676 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_15[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_678 (Activation)     (None, 7, 7, 672)    0           adjust_bn_15[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_680 (Activation)     (None, 7, 7, 672)    0           adjust_bn_15[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_682 (Activation)     (None, 7, 7, 672)    0           adjust_bn_15[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_684 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_15[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 7, 7, 672)    468384      activation_676[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 7, 7, 672)    457632      activation_678[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 7, 7, 672)    468384      activation_680[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 7, 7, 672)    457632      activation_682[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 7, 7, 672)    457632      activation_684[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left1_15[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right1_15\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left2_15[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right2_15\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left5_15[\n",
            "__________________________________________________________________________________________________\n",
            "activation_677 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_679 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_681 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_683 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_685 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 7, 7, 672)    468384      activation_677[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 7, 7, 672)    457632      activation_679[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 7, 7, 672)    468384      activation_681[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 7, 7, 672)    457632      activation_683[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 7, 7, 672)    457632      activation_685[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left1_15[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right1_15\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left2_15[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right2_15\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_15 (AveragePooling (None, 7, 7, 672)    0           normal_bn_1_15[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_15 (AveragePooling (None, 7, 7, 672)    0           adjust_bn_15[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_15 (AveragePoolin (None, 7, 7, 672)    0           adjust_bn_15[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left5_15[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_15 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_15 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_15 (Add)           (None, 7, 7, 672)    0           normal_left3_15[0][0]            \n",
            "                                                                 adjust_bn_15[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_15 (Add)           (None, 7, 7, 672)    0           normal_left4_15[0][0]            \n",
            "                                                                 normal_right4_15[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_15 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_15[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_15 (Concatenate)  (None, 7, 7, 4032)   0           adjust_bn_15[0][0]               \n",
            "                                                                 normal_add_1_15[0][0]            \n",
            "                                                                 normal_add_2_15[0][0]            \n",
            "                                                                 normal_add_3_15[0][0]            \n",
            "                                                                 normal_add_4_15[0][0]            \n",
            "                                                                 normal_add_5_15[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_686 (Activation)     (None, 7, 7, 4032)   0           normal_concat_14[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_687 (Activation)     (None, 7, 7, 4032)   0           normal_concat_15[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_16 (Conv (None, 7, 7, 672)    2709504     activation_686[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_16 (Conv2D)       (None, 7, 7, 672)    2709504     activation_687[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_16 (BatchNormalizatio (None, 7, 7, 672)    2688        adjust_conv_projection_16[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_16 (BatchNormalizat (None, 7, 7, 672)    2688        normal_conv_1_16[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_688 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_16[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_690 (Activation)     (None, 7, 7, 672)    0           adjust_bn_16[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_692 (Activation)     (None, 7, 7, 672)    0           adjust_bn_16[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_694 (Activation)     (None, 7, 7, 672)    0           adjust_bn_16[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_696 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_16[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 7, 7, 672)    468384      activation_688[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 7, 7, 672)    457632      activation_690[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 7, 7, 672)    468384      activation_692[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 7, 7, 672)    457632      activation_694[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 7, 7, 672)    457632      activation_696[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left1_16[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right1_16\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left2_16[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right2_16\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left5_16[\n",
            "__________________________________________________________________________________________________\n",
            "activation_689 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_691 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_693 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_695 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_697 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 7, 7, 672)    468384      activation_689[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 7, 7, 672)    457632      activation_691[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 7, 7, 672)    468384      activation_693[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 7, 7, 672)    457632      activation_695[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 7, 7, 672)    457632      activation_697[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left1_16[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right1_16\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left2_16[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right2_16\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_16 (AveragePooling (None, 7, 7, 672)    0           normal_bn_1_16[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_16 (AveragePooling (None, 7, 7, 672)    0           adjust_bn_16[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_16 (AveragePoolin (None, 7, 7, 672)    0           adjust_bn_16[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left5_16[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_16 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_16 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_16 (Add)           (None, 7, 7, 672)    0           normal_left3_16[0][0]            \n",
            "                                                                 adjust_bn_16[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_16 (Add)           (None, 7, 7, 672)    0           normal_left4_16[0][0]            \n",
            "                                                                 normal_right4_16[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_16 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_16[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_16 (Concatenate)  (None, 7, 7, 4032)   0           adjust_bn_16[0][0]               \n",
            "                                                                 normal_add_1_16[0][0]            \n",
            "                                                                 normal_add_2_16[0][0]            \n",
            "                                                                 normal_add_3_16[0][0]            \n",
            "                                                                 normal_add_4_16[0][0]            \n",
            "                                                                 normal_add_5_16[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_698 (Activation)     (None, 7, 7, 4032)   0           normal_concat_15[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_699 (Activation)     (None, 7, 7, 4032)   0           normal_concat_16[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_17 (Conv (None, 7, 7, 672)    2709504     activation_698[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_17 (Conv2D)       (None, 7, 7, 672)    2709504     activation_699[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_17 (BatchNormalizatio (None, 7, 7, 672)    2688        adjust_conv_projection_17[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_17 (BatchNormalizat (None, 7, 7, 672)    2688        normal_conv_1_17[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_700 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_17[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_702 (Activation)     (None, 7, 7, 672)    0           adjust_bn_17[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_704 (Activation)     (None, 7, 7, 672)    0           adjust_bn_17[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_706 (Activation)     (None, 7, 7, 672)    0           adjust_bn_17[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_708 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_17[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 7, 7, 672)    468384      activation_700[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 7, 7, 672)    457632      activation_702[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 7, 7, 672)    468384      activation_704[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 7, 7, 672)    457632      activation_706[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 7, 7, 672)    457632      activation_708[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left1_17[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right1_17\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left2_17[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right2_17\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left5_17[\n",
            "__________________________________________________________________________________________________\n",
            "activation_701 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_703 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_705 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_707 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_709 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 7, 7, 672)    468384      activation_701[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 7, 7, 672)    457632      activation_703[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 7, 7, 672)    468384      activation_705[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 7, 7, 672)    457632      activation_707[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 7, 7, 672)    457632      activation_709[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left1_17[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right1_17\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left2_17[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right2_17\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_17 (AveragePooling (None, 7, 7, 672)    0           normal_bn_1_17[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_17 (AveragePooling (None, 7, 7, 672)    0           adjust_bn_17[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_17 (AveragePoolin (None, 7, 7, 672)    0           adjust_bn_17[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left5_17[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_17 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_17 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_17 (Add)           (None, 7, 7, 672)    0           normal_left3_17[0][0]            \n",
            "                                                                 adjust_bn_17[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_17 (Add)           (None, 7, 7, 672)    0           normal_left4_17[0][0]            \n",
            "                                                                 normal_right4_17[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_17 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_17[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_17 (Concatenate)  (None, 7, 7, 4032)   0           adjust_bn_17[0][0]               \n",
            "                                                                 normal_add_1_17[0][0]            \n",
            "                                                                 normal_add_2_17[0][0]            \n",
            "                                                                 normal_add_3_17[0][0]            \n",
            "                                                                 normal_add_4_17[0][0]            \n",
            "                                                                 normal_add_5_17[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_710 (Activation)     (None, 7, 7, 4032)   0           normal_concat_16[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_711 (Activation)     (None, 7, 7, 4032)   0           normal_concat_17[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "adjust_conv_projection_18 (Conv (None, 7, 7, 672)    2709504     activation_710[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_conv_1_18 (Conv2D)       (None, 7, 7, 672)    2709504     activation_711[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "adjust_bn_18 (BatchNormalizatio (None, 7, 7, 672)    2688        adjust_conv_projection_18[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "normal_bn_1_18 (BatchNormalizat (None, 7, 7, 672)    2688        normal_conv_1_18[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "activation_712 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_18[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "activation_714 (Activation)     (None, 7, 7, 672)    0           adjust_bn_18[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_716 (Activation)     (None, 7, 7, 672)    0           adjust_bn_18[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_718 (Activation)     (None, 7, 7, 672)    0           adjust_bn_18[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "activation_720 (Activation)     (None, 7, 7, 672)    0           normal_bn_1_18[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left1_1 (None, 7, 7, 672)    468384      activation_712[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right1_ (None, 7, 7, 672)    457632      activation_714[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left2_1 (None, 7, 7, 672)    468384      activation_716[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_right2_ (None, 7, 7, 672)    457632      activation_718[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_normal_left5_1 (None, 7, 7, 672)    457632      activation_720[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left1_18[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right1_18\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left2_18[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_1_normal_right2_18\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_1_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_1_normal_left5_18[\n",
            "__________________________________________________________________________________________________\n",
            "activation_713 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left1_\n",
            "__________________________________________________________________________________________________\n",
            "activation_715 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "activation_717 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left2_\n",
            "__________________________________________________________________________________________________\n",
            "activation_719 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "activation_721 (Activation)     (None, 7, 7, 672)    0           separable_conv_1_bn_normal_left5_\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left1_1 (None, 7, 7, 672)    468384      activation_713[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right1_ (None, 7, 7, 672)    457632      activation_715[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left2_1 (None, 7, 7, 672)    468384      activation_717[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_right2_ (None, 7, 7, 672)    457632      activation_719[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_normal_left5_1 (None, 7, 7, 672)    457632      activation_721[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left1_18[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right1_18\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left2_18[\n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_righ (None, 7, 7, 672)    2688        separable_conv_2_normal_right2_18\n",
            "__________________________________________________________________________________________________\n",
            "normal_left3_18 (AveragePooling (None, 7, 7, 672)    0           normal_bn_1_18[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_left4_18 (AveragePooling (None, 7, 7, 672)    0           adjust_bn_18[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_right4_18 (AveragePoolin (None, 7, 7, 672)    0           adjust_bn_18[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "separable_conv_2_bn_normal_left (None, 7, 7, 672)    2688        separable_conv_2_normal_left5_18[\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_1_18 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left1_\n",
            "                                                                 separable_conv_2_bn_normal_right1\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_2_18 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left2_\n",
            "                                                                 separable_conv_2_bn_normal_right2\n",
            "__________________________________________________________________________________________________\n",
            "normal_add_3_18 (Add)           (None, 7, 7, 672)    0           normal_left3_18[0][0]            \n",
            "                                                                 adjust_bn_18[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_4_18 (Add)           (None, 7, 7, 672)    0           normal_left4_18[0][0]            \n",
            "                                                                 normal_right4_18[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "normal_add_5_18 (Add)           (None, 7, 7, 672)    0           separable_conv_2_bn_normal_left5_\n",
            "                                                                 normal_bn_1_18[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "normal_concat_18 (Concatenate)  (None, 7, 7, 4032)   0           adjust_bn_18[0][0]               \n",
            "                                                                 normal_add_1_18[0][0]            \n",
            "                                                                 normal_add_2_18[0][0]            \n",
            "                                                                 normal_add_3_18[0][0]            \n",
            "                                                                 normal_add_4_18[0][0]            \n",
            "                                                                 normal_add_5_18[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "activation_722 (Activation)     (None, 7, 7, 4032)   0           normal_concat_18[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_9 (AveragePoo (None, 1, 1, 4032)   0           activation_722[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "flatten (Flatten)               (None, 4032)         0           average_pooling2d_9[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "dense_12 (Dense)                (None, 64)           258112      flatten[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout_8 (Dropout)             (None, 64)           0           dense_12[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "dense_13 (Dense)                (None, 2)            130         dropout_8[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 85,175,060\n",
            "Trainable params: 258,242\n",
            "Non-trainable params: 84,916,818\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fb_-jhei4p9k",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(loss=\"categorical_crossentropy\", \n",
        "              optimizer=Nadam(),\n",
        "              metrics=[\"accuracy\"])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H7db8mca4PRP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "316f3dda-f176-4a48-aad0-5ad8b4fa27f5"
      },
      "source": [
        "H3 = model.fit(train_augmented_data, \n",
        "                        steps_per_epoch=8,\n",
        "                        validation_data=validation_data,\n",
        "                        validation_steps=8,\n",
        "                        epochs=50)"
      ],
      "execution_count": 77,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "8/8 [==============================] - 10s 1s/step - loss: 0.5383 - accuracy: 0.7839 - val_loss: 0.2000 - val_accuracy: 0.9625\n",
            "Epoch 2/50\n",
            "8/8 [==============================] - 5s 654ms/step - loss: 0.2293 - accuracy: 0.9110 - val_loss: 0.1665 - val_accuracy: 0.9625\n",
            "Epoch 3/50\n",
            "8/8 [==============================] - 6s 692ms/step - loss: 0.1545 - accuracy: 0.9492 - val_loss: 0.1722 - val_accuracy: 0.9625\n",
            "Epoch 4/50\n",
            "8/8 [==============================] - 5s 653ms/step - loss: 0.1144 - accuracy: 0.9534 - val_loss: 0.1986 - val_accuracy: 0.9625\n",
            "Epoch 5/50\n",
            "8/8 [==============================] - 5s 650ms/step - loss: 0.1597 - accuracy: 0.9492 - val_loss: 0.2031 - val_accuracy: 0.9625\n",
            "Epoch 6/50\n",
            "8/8 [==============================] - 5s 631ms/step - loss: 0.1870 - accuracy: 0.9280 - val_loss: 0.1318 - val_accuracy: 0.9500\n",
            "Epoch 7/50\n",
            "8/8 [==============================] - 5s 657ms/step - loss: 0.1095 - accuracy: 0.9703 - val_loss: 0.1843 - val_accuracy: 0.9625\n",
            "Epoch 8/50\n",
            "8/8 [==============================] - 5s 654ms/step - loss: 0.1074 - accuracy: 0.9619 - val_loss: 0.2370 - val_accuracy: 0.9750\n",
            "Epoch 9/50\n",
            "8/8 [==============================] - 5s 652ms/step - loss: 0.1437 - accuracy: 0.9576 - val_loss: 0.1296 - val_accuracy: 0.9500\n",
            "Epoch 10/50\n",
            "8/8 [==============================] - 5s 656ms/step - loss: 0.1113 - accuracy: 0.9703 - val_loss: 0.2140 - val_accuracy: 0.9750\n",
            "Epoch 11/50\n",
            "8/8 [==============================] - 5s 627ms/step - loss: 0.1013 - accuracy: 0.9746 - val_loss: 0.1414 - val_accuracy: 0.9750\n",
            "Epoch 12/50\n",
            "8/8 [==============================] - 6s 689ms/step - loss: 0.1081 - accuracy: 0.9619 - val_loss: 0.1457 - val_accuracy: 0.9750\n",
            "Epoch 13/50\n",
            "8/8 [==============================] - 5s 658ms/step - loss: 0.1035 - accuracy: 0.9619 - val_loss: 0.2026 - val_accuracy: 0.9750\n",
            "Epoch 14/50\n",
            "8/8 [==============================] - 5s 633ms/step - loss: 0.0512 - accuracy: 0.9831 - val_loss: 0.2340 - val_accuracy: 0.9750\n",
            "Epoch 15/50\n",
            "8/8 [==============================] - 6s 705ms/step - loss: 0.0639 - accuracy: 0.9788 - val_loss: 0.2052 - val_accuracy: 0.9750\n",
            "Epoch 16/50\n",
            "8/8 [==============================] - 5s 658ms/step - loss: 0.0718 - accuracy: 0.9661 - val_loss: 0.1729 - val_accuracy: 0.9750\n",
            "Epoch 17/50\n",
            "8/8 [==============================] - 5s 653ms/step - loss: 0.1023 - accuracy: 0.9492 - val_loss: 0.1642 - val_accuracy: 0.9625\n",
            "Epoch 18/50\n",
            "8/8 [==============================] - 6s 702ms/step - loss: 0.0936 - accuracy: 0.9619 - val_loss: 0.1541 - val_accuracy: 0.9375\n",
            "Epoch 19/50\n",
            "8/8 [==============================] - 5s 662ms/step - loss: 0.0993 - accuracy: 0.9746 - val_loss: 0.1980 - val_accuracy: 0.9750\n",
            "Epoch 20/50\n",
            "8/8 [==============================] - 5s 659ms/step - loss: 0.0491 - accuracy: 0.9831 - val_loss: 0.1975 - val_accuracy: 0.9750\n",
            "Epoch 21/50\n",
            "8/8 [==============================] - 5s 659ms/step - loss: 0.1032 - accuracy: 0.9619 - val_loss: 0.1938 - val_accuracy: 0.9750\n",
            "Epoch 22/50\n",
            "8/8 [==============================] - 5s 653ms/step - loss: 0.0999 - accuracy: 0.9661 - val_loss: 0.1408 - val_accuracy: 0.9625\n",
            "Epoch 23/50\n",
            "8/8 [==============================] - 5s 652ms/step - loss: 0.1040 - accuracy: 0.9703 - val_loss: 0.1611 - val_accuracy: 0.9750\n",
            "Epoch 24/50\n",
            "8/8 [==============================] - 5s 654ms/step - loss: 0.0623 - accuracy: 0.9788 - val_loss: 0.1799 - val_accuracy: 0.9750\n",
            "Epoch 25/50\n",
            "8/8 [==============================] - 6s 697ms/step - loss: 0.0522 - accuracy: 0.9788 - val_loss: 0.1482 - val_accuracy: 0.9625\n",
            "Epoch 26/50\n",
            "8/8 [==============================] - 5s 651ms/step - loss: 0.0552 - accuracy: 0.9915 - val_loss: 0.1828 - val_accuracy: 0.9750\n",
            "Epoch 27/50\n",
            "8/8 [==============================] - 5s 634ms/step - loss: 0.0341 - accuracy: 0.9958 - val_loss: 0.2162 - val_accuracy: 0.9750\n",
            "Epoch 28/50\n",
            "8/8 [==============================] - 5s 653ms/step - loss: 0.0741 - accuracy: 0.9661 - val_loss: 0.1999 - val_accuracy: 0.9750\n",
            "Epoch 29/50\n",
            "8/8 [==============================] - 5s 658ms/step - loss: 0.0657 - accuracy: 0.9703 - val_loss: 0.1754 - val_accuracy: 0.9750\n",
            "Epoch 30/50\n",
            "8/8 [==============================] - 5s 656ms/step - loss: 0.0245 - accuracy: 0.9915 - val_loss: 0.2418 - val_accuracy: 0.9750\n",
            "Epoch 31/50\n",
            "8/8 [==============================] - 6s 705ms/step - loss: 0.0598 - accuracy: 0.9788 - val_loss: 0.1873 - val_accuracy: 0.9750\n",
            "Epoch 32/50\n",
            "8/8 [==============================] - 6s 698ms/step - loss: 0.0878 - accuracy: 0.9661 - val_loss: 0.1850 - val_accuracy: 0.9750\n",
            "Epoch 33/50\n",
            "8/8 [==============================] - 5s 635ms/step - loss: 0.0762 - accuracy: 0.9831 - val_loss: 0.1500 - val_accuracy: 0.9625\n",
            "Epoch 34/50\n",
            "8/8 [==============================] - 6s 705ms/step - loss: 0.0482 - accuracy: 0.9873 - val_loss: 0.1980 - val_accuracy: 0.9750\n",
            "Epoch 35/50\n",
            "8/8 [==============================] - 6s 700ms/step - loss: 0.0272 - accuracy: 0.9873 - val_loss: 0.2090 - val_accuracy: 0.9750\n",
            "Epoch 36/50\n",
            "8/8 [==============================] - 5s 655ms/step - loss: 0.0577 - accuracy: 0.9788 - val_loss: 0.2312 - val_accuracy: 0.9750\n",
            "Epoch 37/50\n",
            "8/8 [==============================] - 5s 659ms/step - loss: 0.0699 - accuracy: 0.9661 - val_loss: 0.2286 - val_accuracy: 0.9750\n",
            "Epoch 38/50\n",
            "8/8 [==============================] - 5s 661ms/step - loss: 0.0488 - accuracy: 0.9915 - val_loss: 0.1827 - val_accuracy: 0.9750\n",
            "Epoch 39/50\n",
            "8/8 [==============================] - 5s 656ms/step - loss: 0.0256 - accuracy: 0.9958 - val_loss: 0.2114 - val_accuracy: 0.9750\n",
            "Epoch 40/50\n",
            "8/8 [==============================] - 5s 635ms/step - loss: 0.0313 - accuracy: 0.9873 - val_loss: 0.1908 - val_accuracy: 0.9750\n",
            "Epoch 41/50\n",
            "8/8 [==============================] - 5s 658ms/step - loss: 0.0991 - accuracy: 0.9661 - val_loss: 0.1628 - val_accuracy: 0.9750\n",
            "Epoch 42/50\n",
            "8/8 [==============================] - 5s 660ms/step - loss: 0.0685 - accuracy: 0.9703 - val_loss: 0.1794 - val_accuracy: 0.9750\n",
            "Epoch 43/50\n",
            "8/8 [==============================] - 5s 642ms/step - loss: 0.0466 - accuracy: 0.9873 - val_loss: 0.1867 - val_accuracy: 0.9750\n",
            "Epoch 44/50\n",
            "8/8 [==============================] - 5s 655ms/step - loss: 0.0417 - accuracy: 0.9873 - val_loss: 0.2064 - val_accuracy: 0.9750\n",
            "Epoch 45/50\n",
            "8/8 [==============================] - 5s 652ms/step - loss: 0.0359 - accuracy: 0.9915 - val_loss: 0.1498 - val_accuracy: 0.9750\n",
            "Epoch 46/50\n",
            "8/8 [==============================] - 5s 655ms/step - loss: 0.0558 - accuracy: 0.9788 - val_loss: 0.1925 - val_accuracy: 0.9750\n",
            "Epoch 47/50\n",
            "8/8 [==============================] - 6s 698ms/step - loss: 0.0501 - accuracy: 0.9746 - val_loss: 0.1882 - val_accuracy: 0.9750\n",
            "Epoch 48/50\n",
            "8/8 [==============================] - 5s 651ms/step - loss: 0.0676 - accuracy: 0.9746 - val_loss: 0.1436 - val_accuracy: 0.9750\n",
            "Epoch 49/50\n",
            "8/8 [==============================] - 5s 653ms/step - loss: 0.0704 - accuracy: 0.9788 - val_loss: 0.1559 - val_accuracy: 0.9750\n",
            "Epoch 50/50\n",
            "8/8 [==============================] - 5s 654ms/step - loss: 0.0491 - accuracy: 0.9873 - val_loss: 0.1603 - val_accuracy: 0.9750\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rzao_y5y8L30",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "outputId": "6b87ea19-2437-49a7-99c3-bc406716e18a"
      },
      "source": [
        "plt.plot(H3.history['accuracy'], label='accuracy')\n",
        "plt.plot(H3.history['val_accuracy'], label = 'val_accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='lower right')"
      ],
      "execution_count": 78,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f979fd9acc0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 78
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qioK-IF8TI66",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "outputId": "33318831-09f2-4943-a1d1-8427b47a15e9"
      },
      "source": [
        "plt.plot(H3.history['loss'], label='loss')\n",
        "plt.plot(H3.history['val_loss'], label = 'val_loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('loss')\n",
        "plt.ylim([0., 1])\n",
        "plt.legend(loc='upper right')"
      ],
      "execution_count": 79,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f979fd02f98>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 79
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A4jaz_J671Gb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "a54f34c0-f004-4cf4-bd0f-a669ff6d421b"
      },
      "source": [
        "predictions = model.predict(test_data,\n",
        "                     batch_size=10,\n",
        "                     verbose=1)\n",
        "y_pred = np.argmax(predictions, axis=1)\n",
        "print(confusion_matrix(test_data.classes, y_pred))\n",
        "print(classification_report(test_data.classes, y_pred, target_names = ['covid', 'healthy']))"
      ],
      "execution_count": 80,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "8/8 [==============================] - 1s 154ms/step\n",
            "[[38  2]\n",
            " [ 0 40]]\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "       covid       1.00      0.95      0.97        40\n",
            "     healthy       0.95      1.00      0.98        40\n",
            "\n",
            "    accuracy                           0.97        80\n",
            "   macro avg       0.98      0.97      0.97        80\n",
            "weighted avg       0.98      0.97      0.97        80\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MWpPDuWxS7CZ",
        "colab_type": "text"
      },
      "source": [
        "### InceptionResNetV2 Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-S3s2XCITJcE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.applications import InceptionResNetV2"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H6_TCE8XUDqP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "baseModel = InceptionResNetV2(weights=\"imagenet\", include_top=False, \n",
        "                  input_tensor=Input(shape=(224, 224, 3)))\n",
        "\n",
        "headModel = baseModel.output\n",
        "headModel = AveragePooling2D(pool_size=(4, 4))(headModel)\n",
        "headModel = Flatten(name=\"flatten\")(headModel)\n",
        "headModel = Dense(64, activation=\"relu\")(headModel)\n",
        "headModel = Dropout(0.5)(headModel)\n",
        "headModel = Dense(2, activation=\"softmax\")(headModel)\n",
        "\n",
        "model = Model(inputs=baseModel.input, outputs=headModel)\n",
        "\n",
        "for layer in baseModel.layers:\n",
        "  layer.trainable = False"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kmoOCwt19-Ei",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "17d36c8c-22f2-48d5-832e-04867f5bf28a"
      },
      "source": [
        "stringlist = []\n",
        "model.summary(print_fn=lambda x: stringlist.append(x))\n",
        "short_model_summary = \"\\n\".join(stringlist)\n",
        "print(short_model_summary)"
      ],
      "execution_count": 91,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_6\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_7 (InputLayer)            [(None, 224, 224, 3) 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_442 (Conv2D)             (None, 111, 111, 32) 864         input_7[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_408 (BatchN (None, 111, 111, 32) 96          conv2d_442[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_926 (Activation)     (None, 111, 111, 32) 0           batch_normalization_408[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_443 (Conv2D)             (None, 109, 109, 32) 9216        activation_926[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_409 (BatchN (None, 109, 109, 32) 96          conv2d_443[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_927 (Activation)     (None, 109, 109, 32) 0           batch_normalization_409[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_444 (Conv2D)             (None, 109, 109, 64) 18432       activation_927[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_410 (BatchN (None, 109, 109, 64) 192         conv2d_444[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_928 (Activation)     (None, 109, 109, 64) 0           batch_normalization_410[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_10 (MaxPooling2D) (None, 54, 54, 64)   0           activation_928[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_445 (Conv2D)             (None, 54, 54, 80)   5120        max_pooling2d_10[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_411 (BatchN (None, 54, 54, 80)   240         conv2d_445[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_929 (Activation)     (None, 54, 54, 80)   0           batch_normalization_411[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_446 (Conv2D)             (None, 52, 52, 192)  138240      activation_929[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_412 (BatchN (None, 52, 52, 192)  576         conv2d_446[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_930 (Activation)     (None, 52, 52, 192)  0           batch_normalization_412[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_11 (MaxPooling2D) (None, 25, 25, 192)  0           activation_930[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_450 (Conv2D)             (None, 25, 25, 64)   12288       max_pooling2d_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_416 (BatchN (None, 25, 25, 64)   192         conv2d_450[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_934 (Activation)     (None, 25, 25, 64)   0           batch_normalization_416[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_448 (Conv2D)             (None, 25, 25, 48)   9216        max_pooling2d_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_451 (Conv2D)             (None, 25, 25, 96)   55296       activation_934[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_414 (BatchN (None, 25, 25, 48)   144         conv2d_448[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_417 (BatchN (None, 25, 25, 96)   288         conv2d_451[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_932 (Activation)     (None, 25, 25, 48)   0           batch_normalization_414[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_935 (Activation)     (None, 25, 25, 96)   0           batch_normalization_417[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_12 (AveragePo (None, 25, 25, 192)  0           max_pooling2d_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_447 (Conv2D)             (None, 25, 25, 96)   18432       max_pooling2d_11[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_449 (Conv2D)             (None, 25, 25, 64)   76800       activation_932[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_452 (Conv2D)             (None, 25, 25, 96)   82944       activation_935[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_453 (Conv2D)             (None, 25, 25, 64)   12288       average_pooling2d_12[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_413 (BatchN (None, 25, 25, 96)   288         conv2d_447[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_415 (BatchN (None, 25, 25, 64)   192         conv2d_449[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_418 (BatchN (None, 25, 25, 96)   288         conv2d_452[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_419 (BatchN (None, 25, 25, 64)   192         conv2d_453[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_931 (Activation)     (None, 25, 25, 96)   0           batch_normalization_413[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_933 (Activation)     (None, 25, 25, 64)   0           batch_normalization_415[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_936 (Activation)     (None, 25, 25, 96)   0           batch_normalization_418[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_937 (Activation)     (None, 25, 25, 64)   0           batch_normalization_419[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "mixed_5b (Concatenate)          (None, 25, 25, 320)  0           activation_931[0][0]             \n",
            "                                                                 activation_933[0][0]             \n",
            "                                                                 activation_936[0][0]             \n",
            "                                                                 activation_937[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_457 (Conv2D)             (None, 25, 25, 32)   10240       mixed_5b[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_423 (BatchN (None, 25, 25, 32)   96          conv2d_457[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_941 (Activation)     (None, 25, 25, 32)   0           batch_normalization_423[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_455 (Conv2D)             (None, 25, 25, 32)   10240       mixed_5b[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_458 (Conv2D)             (None, 25, 25, 48)   13824       activation_941[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_421 (BatchN (None, 25, 25, 32)   96          conv2d_455[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_424 (BatchN (None, 25, 25, 48)   144         conv2d_458[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_939 (Activation)     (None, 25, 25, 32)   0           batch_normalization_421[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_942 (Activation)     (None, 25, 25, 48)   0           batch_normalization_424[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_454 (Conv2D)             (None, 25, 25, 32)   10240       mixed_5b[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_456 (Conv2D)             (None, 25, 25, 32)   9216        activation_939[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_459 (Conv2D)             (None, 25, 25, 64)   27648       activation_942[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_420 (BatchN (None, 25, 25, 32)   96          conv2d_454[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_422 (BatchN (None, 25, 25, 32)   96          conv2d_456[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_425 (BatchN (None, 25, 25, 64)   192         conv2d_459[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_938 (Activation)     (None, 25, 25, 32)   0           batch_normalization_420[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_940 (Activation)     (None, 25, 25, 32)   0           batch_normalization_422[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_943 (Activation)     (None, 25, 25, 64)   0           batch_normalization_425[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_1_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_938[0][0]             \n",
            "                                                                 activation_940[0][0]             \n",
            "                                                                 activation_943[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_1_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_1_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_1 (Lambda)              (None, 25, 25, 320)  0           mixed_5b[0][0]                   \n",
            "                                                                 block35_1_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_1_ac (Activation)       (None, 25, 25, 320)  0           block35_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_463 (Conv2D)             (None, 25, 25, 32)   10240       block35_1_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_429 (BatchN (None, 25, 25, 32)   96          conv2d_463[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_947 (Activation)     (None, 25, 25, 32)   0           batch_normalization_429[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_461 (Conv2D)             (None, 25, 25, 32)   10240       block35_1_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_464 (Conv2D)             (None, 25, 25, 48)   13824       activation_947[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_427 (BatchN (None, 25, 25, 32)   96          conv2d_461[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_430 (BatchN (None, 25, 25, 48)   144         conv2d_464[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_945 (Activation)     (None, 25, 25, 32)   0           batch_normalization_427[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_948 (Activation)     (None, 25, 25, 48)   0           batch_normalization_430[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_460 (Conv2D)             (None, 25, 25, 32)   10240       block35_1_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_462 (Conv2D)             (None, 25, 25, 32)   9216        activation_945[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_465 (Conv2D)             (None, 25, 25, 64)   27648       activation_948[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_426 (BatchN (None, 25, 25, 32)   96          conv2d_460[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_428 (BatchN (None, 25, 25, 32)   96          conv2d_462[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_431 (BatchN (None, 25, 25, 64)   192         conv2d_465[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_944 (Activation)     (None, 25, 25, 32)   0           batch_normalization_426[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_946 (Activation)     (None, 25, 25, 32)   0           batch_normalization_428[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_949 (Activation)     (None, 25, 25, 64)   0           batch_normalization_431[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_2_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_944[0][0]             \n",
            "                                                                 activation_946[0][0]             \n",
            "                                                                 activation_949[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_2_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_2_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_2 (Lambda)              (None, 25, 25, 320)  0           block35_1_ac[0][0]               \n",
            "                                                                 block35_2_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_2_ac (Activation)       (None, 25, 25, 320)  0           block35_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_469 (Conv2D)             (None, 25, 25, 32)   10240       block35_2_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_435 (BatchN (None, 25, 25, 32)   96          conv2d_469[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_953 (Activation)     (None, 25, 25, 32)   0           batch_normalization_435[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_467 (Conv2D)             (None, 25, 25, 32)   10240       block35_2_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_470 (Conv2D)             (None, 25, 25, 48)   13824       activation_953[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_433 (BatchN (None, 25, 25, 32)   96          conv2d_467[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_436 (BatchN (None, 25, 25, 48)   144         conv2d_470[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_951 (Activation)     (None, 25, 25, 32)   0           batch_normalization_433[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_954 (Activation)     (None, 25, 25, 48)   0           batch_normalization_436[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_466 (Conv2D)             (None, 25, 25, 32)   10240       block35_2_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_468 (Conv2D)             (None, 25, 25, 32)   9216        activation_951[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_471 (Conv2D)             (None, 25, 25, 64)   27648       activation_954[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_432 (BatchN (None, 25, 25, 32)   96          conv2d_466[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_434 (BatchN (None, 25, 25, 32)   96          conv2d_468[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_437 (BatchN (None, 25, 25, 64)   192         conv2d_471[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_950 (Activation)     (None, 25, 25, 32)   0           batch_normalization_432[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_952 (Activation)     (None, 25, 25, 32)   0           batch_normalization_434[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_955 (Activation)     (None, 25, 25, 64)   0           batch_normalization_437[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_3_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_950[0][0]             \n",
            "                                                                 activation_952[0][0]             \n",
            "                                                                 activation_955[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_3_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_3_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_3 (Lambda)              (None, 25, 25, 320)  0           block35_2_ac[0][0]               \n",
            "                                                                 block35_3_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_3_ac (Activation)       (None, 25, 25, 320)  0           block35_3[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_475 (Conv2D)             (None, 25, 25, 32)   10240       block35_3_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_441 (BatchN (None, 25, 25, 32)   96          conv2d_475[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_959 (Activation)     (None, 25, 25, 32)   0           batch_normalization_441[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_473 (Conv2D)             (None, 25, 25, 32)   10240       block35_3_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_476 (Conv2D)             (None, 25, 25, 48)   13824       activation_959[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_439 (BatchN (None, 25, 25, 32)   96          conv2d_473[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_442 (BatchN (None, 25, 25, 48)   144         conv2d_476[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_957 (Activation)     (None, 25, 25, 32)   0           batch_normalization_439[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_960 (Activation)     (None, 25, 25, 48)   0           batch_normalization_442[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_472 (Conv2D)             (None, 25, 25, 32)   10240       block35_3_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_474 (Conv2D)             (None, 25, 25, 32)   9216        activation_957[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_477 (Conv2D)             (None, 25, 25, 64)   27648       activation_960[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_438 (BatchN (None, 25, 25, 32)   96          conv2d_472[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_440 (BatchN (None, 25, 25, 32)   96          conv2d_474[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_443 (BatchN (None, 25, 25, 64)   192         conv2d_477[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_956 (Activation)     (None, 25, 25, 32)   0           batch_normalization_438[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_958 (Activation)     (None, 25, 25, 32)   0           batch_normalization_440[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_961 (Activation)     (None, 25, 25, 64)   0           batch_normalization_443[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_4_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_956[0][0]             \n",
            "                                                                 activation_958[0][0]             \n",
            "                                                                 activation_961[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_4_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_4_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_4 (Lambda)              (None, 25, 25, 320)  0           block35_3_ac[0][0]               \n",
            "                                                                 block35_4_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_4_ac (Activation)       (None, 25, 25, 320)  0           block35_4[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_481 (Conv2D)             (None, 25, 25, 32)   10240       block35_4_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_447 (BatchN (None, 25, 25, 32)   96          conv2d_481[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_965 (Activation)     (None, 25, 25, 32)   0           batch_normalization_447[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_479 (Conv2D)             (None, 25, 25, 32)   10240       block35_4_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_482 (Conv2D)             (None, 25, 25, 48)   13824       activation_965[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_445 (BatchN (None, 25, 25, 32)   96          conv2d_479[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_448 (BatchN (None, 25, 25, 48)   144         conv2d_482[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_963 (Activation)     (None, 25, 25, 32)   0           batch_normalization_445[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_966 (Activation)     (None, 25, 25, 48)   0           batch_normalization_448[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_478 (Conv2D)             (None, 25, 25, 32)   10240       block35_4_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_480 (Conv2D)             (None, 25, 25, 32)   9216        activation_963[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_483 (Conv2D)             (None, 25, 25, 64)   27648       activation_966[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_444 (BatchN (None, 25, 25, 32)   96          conv2d_478[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_446 (BatchN (None, 25, 25, 32)   96          conv2d_480[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_449 (BatchN (None, 25, 25, 64)   192         conv2d_483[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_962 (Activation)     (None, 25, 25, 32)   0           batch_normalization_444[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_964 (Activation)     (None, 25, 25, 32)   0           batch_normalization_446[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_967 (Activation)     (None, 25, 25, 64)   0           batch_normalization_449[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_5_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_962[0][0]             \n",
            "                                                                 activation_964[0][0]             \n",
            "                                                                 activation_967[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_5_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_5_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_5 (Lambda)              (None, 25, 25, 320)  0           block35_4_ac[0][0]               \n",
            "                                                                 block35_5_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_5_ac (Activation)       (None, 25, 25, 320)  0           block35_5[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_487 (Conv2D)             (None, 25, 25, 32)   10240       block35_5_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_453 (BatchN (None, 25, 25, 32)   96          conv2d_487[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_971 (Activation)     (None, 25, 25, 32)   0           batch_normalization_453[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_485 (Conv2D)             (None, 25, 25, 32)   10240       block35_5_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_488 (Conv2D)             (None, 25, 25, 48)   13824       activation_971[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_451 (BatchN (None, 25, 25, 32)   96          conv2d_485[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_454 (BatchN (None, 25, 25, 48)   144         conv2d_488[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_969 (Activation)     (None, 25, 25, 32)   0           batch_normalization_451[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_972 (Activation)     (None, 25, 25, 48)   0           batch_normalization_454[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_484 (Conv2D)             (None, 25, 25, 32)   10240       block35_5_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_486 (Conv2D)             (None, 25, 25, 32)   9216        activation_969[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_489 (Conv2D)             (None, 25, 25, 64)   27648       activation_972[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_450 (BatchN (None, 25, 25, 32)   96          conv2d_484[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_452 (BatchN (None, 25, 25, 32)   96          conv2d_486[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_455 (BatchN (None, 25, 25, 64)   192         conv2d_489[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_968 (Activation)     (None, 25, 25, 32)   0           batch_normalization_450[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_970 (Activation)     (None, 25, 25, 32)   0           batch_normalization_452[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_973 (Activation)     (None, 25, 25, 64)   0           batch_normalization_455[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_6_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_968[0][0]             \n",
            "                                                                 activation_970[0][0]             \n",
            "                                                                 activation_973[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_6_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_6_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_6 (Lambda)              (None, 25, 25, 320)  0           block35_5_ac[0][0]               \n",
            "                                                                 block35_6_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_6_ac (Activation)       (None, 25, 25, 320)  0           block35_6[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_493 (Conv2D)             (None, 25, 25, 32)   10240       block35_6_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_459 (BatchN (None, 25, 25, 32)   96          conv2d_493[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_977 (Activation)     (None, 25, 25, 32)   0           batch_normalization_459[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_491 (Conv2D)             (None, 25, 25, 32)   10240       block35_6_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_494 (Conv2D)             (None, 25, 25, 48)   13824       activation_977[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_457 (BatchN (None, 25, 25, 32)   96          conv2d_491[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_460 (BatchN (None, 25, 25, 48)   144         conv2d_494[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_975 (Activation)     (None, 25, 25, 32)   0           batch_normalization_457[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_978 (Activation)     (None, 25, 25, 48)   0           batch_normalization_460[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_490 (Conv2D)             (None, 25, 25, 32)   10240       block35_6_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_492 (Conv2D)             (None, 25, 25, 32)   9216        activation_975[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_495 (Conv2D)             (None, 25, 25, 64)   27648       activation_978[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_456 (BatchN (None, 25, 25, 32)   96          conv2d_490[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_458 (BatchN (None, 25, 25, 32)   96          conv2d_492[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_461 (BatchN (None, 25, 25, 64)   192         conv2d_495[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_974 (Activation)     (None, 25, 25, 32)   0           batch_normalization_456[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_976 (Activation)     (None, 25, 25, 32)   0           batch_normalization_458[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_979 (Activation)     (None, 25, 25, 64)   0           batch_normalization_461[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_7_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_974[0][0]             \n",
            "                                                                 activation_976[0][0]             \n",
            "                                                                 activation_979[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_7_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_7_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_7 (Lambda)              (None, 25, 25, 320)  0           block35_6_ac[0][0]               \n",
            "                                                                 block35_7_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_7_ac (Activation)       (None, 25, 25, 320)  0           block35_7[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_499 (Conv2D)             (None, 25, 25, 32)   10240       block35_7_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_465 (BatchN (None, 25, 25, 32)   96          conv2d_499[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_983 (Activation)     (None, 25, 25, 32)   0           batch_normalization_465[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_497 (Conv2D)             (None, 25, 25, 32)   10240       block35_7_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_500 (Conv2D)             (None, 25, 25, 48)   13824       activation_983[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_463 (BatchN (None, 25, 25, 32)   96          conv2d_497[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_466 (BatchN (None, 25, 25, 48)   144         conv2d_500[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_981 (Activation)     (None, 25, 25, 32)   0           batch_normalization_463[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_984 (Activation)     (None, 25, 25, 48)   0           batch_normalization_466[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_496 (Conv2D)             (None, 25, 25, 32)   10240       block35_7_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_498 (Conv2D)             (None, 25, 25, 32)   9216        activation_981[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_501 (Conv2D)             (None, 25, 25, 64)   27648       activation_984[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_462 (BatchN (None, 25, 25, 32)   96          conv2d_496[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_464 (BatchN (None, 25, 25, 32)   96          conv2d_498[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_467 (BatchN (None, 25, 25, 64)   192         conv2d_501[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_980 (Activation)     (None, 25, 25, 32)   0           batch_normalization_462[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_982 (Activation)     (None, 25, 25, 32)   0           batch_normalization_464[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_985 (Activation)     (None, 25, 25, 64)   0           batch_normalization_467[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_8_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_980[0][0]             \n",
            "                                                                 activation_982[0][0]             \n",
            "                                                                 activation_985[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_8_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_8_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_8 (Lambda)              (None, 25, 25, 320)  0           block35_7_ac[0][0]               \n",
            "                                                                 block35_8_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_8_ac (Activation)       (None, 25, 25, 320)  0           block35_8[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_505 (Conv2D)             (None, 25, 25, 32)   10240       block35_8_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_471 (BatchN (None, 25, 25, 32)   96          conv2d_505[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_989 (Activation)     (None, 25, 25, 32)   0           batch_normalization_471[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_503 (Conv2D)             (None, 25, 25, 32)   10240       block35_8_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_506 (Conv2D)             (None, 25, 25, 48)   13824       activation_989[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_469 (BatchN (None, 25, 25, 32)   96          conv2d_503[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_472 (BatchN (None, 25, 25, 48)   144         conv2d_506[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_987 (Activation)     (None, 25, 25, 32)   0           batch_normalization_469[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_990 (Activation)     (None, 25, 25, 48)   0           batch_normalization_472[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_502 (Conv2D)             (None, 25, 25, 32)   10240       block35_8_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_504 (Conv2D)             (None, 25, 25, 32)   9216        activation_987[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_507 (Conv2D)             (None, 25, 25, 64)   27648       activation_990[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_468 (BatchN (None, 25, 25, 32)   96          conv2d_502[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_470 (BatchN (None, 25, 25, 32)   96          conv2d_504[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_473 (BatchN (None, 25, 25, 64)   192         conv2d_507[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_986 (Activation)     (None, 25, 25, 32)   0           batch_normalization_468[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_988 (Activation)     (None, 25, 25, 32)   0           batch_normalization_470[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_991 (Activation)     (None, 25, 25, 64)   0           batch_normalization_473[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_9_mixed (Concatenate)   (None, 25, 25, 128)  0           activation_986[0][0]             \n",
            "                                                                 activation_988[0][0]             \n",
            "                                                                 activation_991[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_9_conv (Conv2D)         (None, 25, 25, 320)  41280       block35_9_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_9 (Lambda)              (None, 25, 25, 320)  0           block35_8_ac[0][0]               \n",
            "                                                                 block35_9_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_9_ac (Activation)       (None, 25, 25, 320)  0           block35_9[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_511 (Conv2D)             (None, 25, 25, 32)   10240       block35_9_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_477 (BatchN (None, 25, 25, 32)   96          conv2d_511[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_995 (Activation)     (None, 25, 25, 32)   0           batch_normalization_477[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_509 (Conv2D)             (None, 25, 25, 32)   10240       block35_9_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_512 (Conv2D)             (None, 25, 25, 48)   13824       activation_995[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_475 (BatchN (None, 25, 25, 32)   96          conv2d_509[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_478 (BatchN (None, 25, 25, 48)   144         conv2d_512[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_993 (Activation)     (None, 25, 25, 32)   0           batch_normalization_475[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_996 (Activation)     (None, 25, 25, 48)   0           batch_normalization_478[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_508 (Conv2D)             (None, 25, 25, 32)   10240       block35_9_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_510 (Conv2D)             (None, 25, 25, 32)   9216        activation_993[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_513 (Conv2D)             (None, 25, 25, 64)   27648       activation_996[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_474 (BatchN (None, 25, 25, 32)   96          conv2d_508[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_476 (BatchN (None, 25, 25, 32)   96          conv2d_510[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_479 (BatchN (None, 25, 25, 64)   192         conv2d_513[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_992 (Activation)     (None, 25, 25, 32)   0           batch_normalization_474[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_994 (Activation)     (None, 25, 25, 32)   0           batch_normalization_476[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_997 (Activation)     (None, 25, 25, 64)   0           batch_normalization_479[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block35_10_mixed (Concatenate)  (None, 25, 25, 128)  0           activation_992[0][0]             \n",
            "                                                                 activation_994[0][0]             \n",
            "                                                                 activation_997[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block35_10_conv (Conv2D)        (None, 25, 25, 320)  41280       block35_10_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block35_10 (Lambda)             (None, 25, 25, 320)  0           block35_9_ac[0][0]               \n",
            "                                                                 block35_10_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block35_10_ac (Activation)      (None, 25, 25, 320)  0           block35_10[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_515 (Conv2D)             (None, 25, 25, 256)  81920       block35_10_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_481 (BatchN (None, 25, 25, 256)  768         conv2d_515[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_999 (Activation)     (None, 25, 25, 256)  0           batch_normalization_481[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_516 (Conv2D)             (None, 25, 25, 256)  589824      activation_999[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_482 (BatchN (None, 25, 25, 256)  768         conv2d_516[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1000 (Activation)    (None, 25, 25, 256)  0           batch_normalization_482[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_514 (Conv2D)             (None, 12, 12, 384)  1105920     block35_10_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_517 (Conv2D)             (None, 12, 12, 384)  884736      activation_1000[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_480 (BatchN (None, 12, 12, 384)  1152        conv2d_514[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_483 (BatchN (None, 12, 12, 384)  1152        conv2d_517[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_998 (Activation)     (None, 12, 12, 384)  0           batch_normalization_480[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1001 (Activation)    (None, 12, 12, 384)  0           batch_normalization_483[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_12 (MaxPooling2D) (None, 12, 12, 320)  0           block35_10_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "mixed_6a (Concatenate)          (None, 12, 12, 1088) 0           activation_998[0][0]             \n",
            "                                                                 activation_1001[0][0]            \n",
            "                                                                 max_pooling2d_12[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_519 (Conv2D)             (None, 12, 12, 128)  139264      mixed_6a[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_485 (BatchN (None, 12, 12, 128)  384         conv2d_519[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1003 (Activation)    (None, 12, 12, 128)  0           batch_normalization_485[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_520 (Conv2D)             (None, 12, 12, 160)  143360      activation_1003[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_486 (BatchN (None, 12, 12, 160)  480         conv2d_520[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1004 (Activation)    (None, 12, 12, 160)  0           batch_normalization_486[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_518 (Conv2D)             (None, 12, 12, 192)  208896      mixed_6a[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_521 (Conv2D)             (None, 12, 12, 192)  215040      activation_1004[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_484 (BatchN (None, 12, 12, 192)  576         conv2d_518[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_487 (BatchN (None, 12, 12, 192)  576         conv2d_521[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1002 (Activation)    (None, 12, 12, 192)  0           batch_normalization_484[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1005 (Activation)    (None, 12, 12, 192)  0           batch_normalization_487[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_1_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1002[0][0]            \n",
            "                                                                 activation_1005[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_1_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_1_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_1 (Lambda)              (None, 12, 12, 1088) 0           mixed_6a[0][0]                   \n",
            "                                                                 block17_1_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_1_ac (Activation)       (None, 12, 12, 1088) 0           block17_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_523 (Conv2D)             (None, 12, 12, 128)  139264      block17_1_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_489 (BatchN (None, 12, 12, 128)  384         conv2d_523[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1007 (Activation)    (None, 12, 12, 128)  0           batch_normalization_489[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_524 (Conv2D)             (None, 12, 12, 160)  143360      activation_1007[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_490 (BatchN (None, 12, 12, 160)  480         conv2d_524[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1008 (Activation)    (None, 12, 12, 160)  0           batch_normalization_490[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_522 (Conv2D)             (None, 12, 12, 192)  208896      block17_1_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_525 (Conv2D)             (None, 12, 12, 192)  215040      activation_1008[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_488 (BatchN (None, 12, 12, 192)  576         conv2d_522[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_491 (BatchN (None, 12, 12, 192)  576         conv2d_525[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1006 (Activation)    (None, 12, 12, 192)  0           batch_normalization_488[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1009 (Activation)    (None, 12, 12, 192)  0           batch_normalization_491[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_2_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1006[0][0]            \n",
            "                                                                 activation_1009[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_2_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_2_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_2 (Lambda)              (None, 12, 12, 1088) 0           block17_1_ac[0][0]               \n",
            "                                                                 block17_2_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_2_ac (Activation)       (None, 12, 12, 1088) 0           block17_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_527 (Conv2D)             (None, 12, 12, 128)  139264      block17_2_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_493 (BatchN (None, 12, 12, 128)  384         conv2d_527[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1011 (Activation)    (None, 12, 12, 128)  0           batch_normalization_493[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_528 (Conv2D)             (None, 12, 12, 160)  143360      activation_1011[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_494 (BatchN (None, 12, 12, 160)  480         conv2d_528[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1012 (Activation)    (None, 12, 12, 160)  0           batch_normalization_494[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_526 (Conv2D)             (None, 12, 12, 192)  208896      block17_2_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_529 (Conv2D)             (None, 12, 12, 192)  215040      activation_1012[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_492 (BatchN (None, 12, 12, 192)  576         conv2d_526[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_495 (BatchN (None, 12, 12, 192)  576         conv2d_529[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1010 (Activation)    (None, 12, 12, 192)  0           batch_normalization_492[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1013 (Activation)    (None, 12, 12, 192)  0           batch_normalization_495[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_3_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1010[0][0]            \n",
            "                                                                 activation_1013[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_3_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_3_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_3 (Lambda)              (None, 12, 12, 1088) 0           block17_2_ac[0][0]               \n",
            "                                                                 block17_3_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_3_ac (Activation)       (None, 12, 12, 1088) 0           block17_3[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_531 (Conv2D)             (None, 12, 12, 128)  139264      block17_3_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_497 (BatchN (None, 12, 12, 128)  384         conv2d_531[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1015 (Activation)    (None, 12, 12, 128)  0           batch_normalization_497[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_532 (Conv2D)             (None, 12, 12, 160)  143360      activation_1015[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_498 (BatchN (None, 12, 12, 160)  480         conv2d_532[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1016 (Activation)    (None, 12, 12, 160)  0           batch_normalization_498[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_530 (Conv2D)             (None, 12, 12, 192)  208896      block17_3_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_533 (Conv2D)             (None, 12, 12, 192)  215040      activation_1016[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_496 (BatchN (None, 12, 12, 192)  576         conv2d_530[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_499 (BatchN (None, 12, 12, 192)  576         conv2d_533[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1014 (Activation)    (None, 12, 12, 192)  0           batch_normalization_496[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1017 (Activation)    (None, 12, 12, 192)  0           batch_normalization_499[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_4_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1014[0][0]            \n",
            "                                                                 activation_1017[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_4_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_4_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_4 (Lambda)              (None, 12, 12, 1088) 0           block17_3_ac[0][0]               \n",
            "                                                                 block17_4_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_4_ac (Activation)       (None, 12, 12, 1088) 0           block17_4[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_535 (Conv2D)             (None, 12, 12, 128)  139264      block17_4_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_501 (BatchN (None, 12, 12, 128)  384         conv2d_535[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1019 (Activation)    (None, 12, 12, 128)  0           batch_normalization_501[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_536 (Conv2D)             (None, 12, 12, 160)  143360      activation_1019[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_502 (BatchN (None, 12, 12, 160)  480         conv2d_536[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1020 (Activation)    (None, 12, 12, 160)  0           batch_normalization_502[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_534 (Conv2D)             (None, 12, 12, 192)  208896      block17_4_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_537 (Conv2D)             (None, 12, 12, 192)  215040      activation_1020[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_500 (BatchN (None, 12, 12, 192)  576         conv2d_534[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_503 (BatchN (None, 12, 12, 192)  576         conv2d_537[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1018 (Activation)    (None, 12, 12, 192)  0           batch_normalization_500[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1021 (Activation)    (None, 12, 12, 192)  0           batch_normalization_503[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_5_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1018[0][0]            \n",
            "                                                                 activation_1021[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_5_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_5_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_5 (Lambda)              (None, 12, 12, 1088) 0           block17_4_ac[0][0]               \n",
            "                                                                 block17_5_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_5_ac (Activation)       (None, 12, 12, 1088) 0           block17_5[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_539 (Conv2D)             (None, 12, 12, 128)  139264      block17_5_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_505 (BatchN (None, 12, 12, 128)  384         conv2d_539[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1023 (Activation)    (None, 12, 12, 128)  0           batch_normalization_505[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_540 (Conv2D)             (None, 12, 12, 160)  143360      activation_1023[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_506 (BatchN (None, 12, 12, 160)  480         conv2d_540[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1024 (Activation)    (None, 12, 12, 160)  0           batch_normalization_506[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_538 (Conv2D)             (None, 12, 12, 192)  208896      block17_5_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_541 (Conv2D)             (None, 12, 12, 192)  215040      activation_1024[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_504 (BatchN (None, 12, 12, 192)  576         conv2d_538[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_507 (BatchN (None, 12, 12, 192)  576         conv2d_541[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1022 (Activation)    (None, 12, 12, 192)  0           batch_normalization_504[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1025 (Activation)    (None, 12, 12, 192)  0           batch_normalization_507[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_6_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1022[0][0]            \n",
            "                                                                 activation_1025[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_6_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_6_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_6 (Lambda)              (None, 12, 12, 1088) 0           block17_5_ac[0][0]               \n",
            "                                                                 block17_6_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_6_ac (Activation)       (None, 12, 12, 1088) 0           block17_6[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_543 (Conv2D)             (None, 12, 12, 128)  139264      block17_6_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_509 (BatchN (None, 12, 12, 128)  384         conv2d_543[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1027 (Activation)    (None, 12, 12, 128)  0           batch_normalization_509[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_544 (Conv2D)             (None, 12, 12, 160)  143360      activation_1027[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_510 (BatchN (None, 12, 12, 160)  480         conv2d_544[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1028 (Activation)    (None, 12, 12, 160)  0           batch_normalization_510[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_542 (Conv2D)             (None, 12, 12, 192)  208896      block17_6_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_545 (Conv2D)             (None, 12, 12, 192)  215040      activation_1028[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_508 (BatchN (None, 12, 12, 192)  576         conv2d_542[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_511 (BatchN (None, 12, 12, 192)  576         conv2d_545[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1026 (Activation)    (None, 12, 12, 192)  0           batch_normalization_508[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1029 (Activation)    (None, 12, 12, 192)  0           batch_normalization_511[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_7_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1026[0][0]            \n",
            "                                                                 activation_1029[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_7_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_7_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_7 (Lambda)              (None, 12, 12, 1088) 0           block17_6_ac[0][0]               \n",
            "                                                                 block17_7_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_7_ac (Activation)       (None, 12, 12, 1088) 0           block17_7[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_547 (Conv2D)             (None, 12, 12, 128)  139264      block17_7_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_513 (BatchN (None, 12, 12, 128)  384         conv2d_547[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1031 (Activation)    (None, 12, 12, 128)  0           batch_normalization_513[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_548 (Conv2D)             (None, 12, 12, 160)  143360      activation_1031[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_514 (BatchN (None, 12, 12, 160)  480         conv2d_548[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1032 (Activation)    (None, 12, 12, 160)  0           batch_normalization_514[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_546 (Conv2D)             (None, 12, 12, 192)  208896      block17_7_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_549 (Conv2D)             (None, 12, 12, 192)  215040      activation_1032[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_512 (BatchN (None, 12, 12, 192)  576         conv2d_546[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_515 (BatchN (None, 12, 12, 192)  576         conv2d_549[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1030 (Activation)    (None, 12, 12, 192)  0           batch_normalization_512[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1033 (Activation)    (None, 12, 12, 192)  0           batch_normalization_515[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_8_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1030[0][0]            \n",
            "                                                                 activation_1033[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_8_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_8_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_8 (Lambda)              (None, 12, 12, 1088) 0           block17_7_ac[0][0]               \n",
            "                                                                 block17_8_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_8_ac (Activation)       (None, 12, 12, 1088) 0           block17_8[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_551 (Conv2D)             (None, 12, 12, 128)  139264      block17_8_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_517 (BatchN (None, 12, 12, 128)  384         conv2d_551[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1035 (Activation)    (None, 12, 12, 128)  0           batch_normalization_517[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_552 (Conv2D)             (None, 12, 12, 160)  143360      activation_1035[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_518 (BatchN (None, 12, 12, 160)  480         conv2d_552[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1036 (Activation)    (None, 12, 12, 160)  0           batch_normalization_518[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_550 (Conv2D)             (None, 12, 12, 192)  208896      block17_8_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_553 (Conv2D)             (None, 12, 12, 192)  215040      activation_1036[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_516 (BatchN (None, 12, 12, 192)  576         conv2d_550[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_519 (BatchN (None, 12, 12, 192)  576         conv2d_553[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1034 (Activation)    (None, 12, 12, 192)  0           batch_normalization_516[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1037 (Activation)    (None, 12, 12, 192)  0           batch_normalization_519[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_9_mixed (Concatenate)   (None, 12, 12, 384)  0           activation_1034[0][0]            \n",
            "                                                                 activation_1037[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_9_conv (Conv2D)         (None, 12, 12, 1088) 418880      block17_9_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_9 (Lambda)              (None, 12, 12, 1088) 0           block17_8_ac[0][0]               \n",
            "                                                                 block17_9_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block17_9_ac (Activation)       (None, 12, 12, 1088) 0           block17_9[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_555 (Conv2D)             (None, 12, 12, 128)  139264      block17_9_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_521 (BatchN (None, 12, 12, 128)  384         conv2d_555[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1039 (Activation)    (None, 12, 12, 128)  0           batch_normalization_521[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_556 (Conv2D)             (None, 12, 12, 160)  143360      activation_1039[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_522 (BatchN (None, 12, 12, 160)  480         conv2d_556[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1040 (Activation)    (None, 12, 12, 160)  0           batch_normalization_522[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_554 (Conv2D)             (None, 12, 12, 192)  208896      block17_9_ac[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_557 (Conv2D)             (None, 12, 12, 192)  215040      activation_1040[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_520 (BatchN (None, 12, 12, 192)  576         conv2d_554[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_523 (BatchN (None, 12, 12, 192)  576         conv2d_557[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1038 (Activation)    (None, 12, 12, 192)  0           batch_normalization_520[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1041 (Activation)    (None, 12, 12, 192)  0           batch_normalization_523[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_10_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1038[0][0]            \n",
            "                                                                 activation_1041[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_10_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_10_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_10 (Lambda)             (None, 12, 12, 1088) 0           block17_9_ac[0][0]               \n",
            "                                                                 block17_10_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_10_ac (Activation)      (None, 12, 12, 1088) 0           block17_10[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_559 (Conv2D)             (None, 12, 12, 128)  139264      block17_10_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_525 (BatchN (None, 12, 12, 128)  384         conv2d_559[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1043 (Activation)    (None, 12, 12, 128)  0           batch_normalization_525[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_560 (Conv2D)             (None, 12, 12, 160)  143360      activation_1043[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_526 (BatchN (None, 12, 12, 160)  480         conv2d_560[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1044 (Activation)    (None, 12, 12, 160)  0           batch_normalization_526[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_558 (Conv2D)             (None, 12, 12, 192)  208896      block17_10_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_561 (Conv2D)             (None, 12, 12, 192)  215040      activation_1044[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_524 (BatchN (None, 12, 12, 192)  576         conv2d_558[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_527 (BatchN (None, 12, 12, 192)  576         conv2d_561[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1042 (Activation)    (None, 12, 12, 192)  0           batch_normalization_524[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1045 (Activation)    (None, 12, 12, 192)  0           batch_normalization_527[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_11_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1042[0][0]            \n",
            "                                                                 activation_1045[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_11_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_11_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_11 (Lambda)             (None, 12, 12, 1088) 0           block17_10_ac[0][0]              \n",
            "                                                                 block17_11_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_11_ac (Activation)      (None, 12, 12, 1088) 0           block17_11[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_563 (Conv2D)             (None, 12, 12, 128)  139264      block17_11_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_529 (BatchN (None, 12, 12, 128)  384         conv2d_563[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1047 (Activation)    (None, 12, 12, 128)  0           batch_normalization_529[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_564 (Conv2D)             (None, 12, 12, 160)  143360      activation_1047[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_530 (BatchN (None, 12, 12, 160)  480         conv2d_564[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1048 (Activation)    (None, 12, 12, 160)  0           batch_normalization_530[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_562 (Conv2D)             (None, 12, 12, 192)  208896      block17_11_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_565 (Conv2D)             (None, 12, 12, 192)  215040      activation_1048[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_528 (BatchN (None, 12, 12, 192)  576         conv2d_562[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_531 (BatchN (None, 12, 12, 192)  576         conv2d_565[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1046 (Activation)    (None, 12, 12, 192)  0           batch_normalization_528[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1049 (Activation)    (None, 12, 12, 192)  0           batch_normalization_531[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_12_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1046[0][0]            \n",
            "                                                                 activation_1049[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_12_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_12_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_12 (Lambda)             (None, 12, 12, 1088) 0           block17_11_ac[0][0]              \n",
            "                                                                 block17_12_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_12_ac (Activation)      (None, 12, 12, 1088) 0           block17_12[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_567 (Conv2D)             (None, 12, 12, 128)  139264      block17_12_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_533 (BatchN (None, 12, 12, 128)  384         conv2d_567[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1051 (Activation)    (None, 12, 12, 128)  0           batch_normalization_533[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_568 (Conv2D)             (None, 12, 12, 160)  143360      activation_1051[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_534 (BatchN (None, 12, 12, 160)  480         conv2d_568[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1052 (Activation)    (None, 12, 12, 160)  0           batch_normalization_534[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_566 (Conv2D)             (None, 12, 12, 192)  208896      block17_12_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_569 (Conv2D)             (None, 12, 12, 192)  215040      activation_1052[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_532 (BatchN (None, 12, 12, 192)  576         conv2d_566[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_535 (BatchN (None, 12, 12, 192)  576         conv2d_569[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1050 (Activation)    (None, 12, 12, 192)  0           batch_normalization_532[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1053 (Activation)    (None, 12, 12, 192)  0           batch_normalization_535[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_13_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1050[0][0]            \n",
            "                                                                 activation_1053[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_13_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_13_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_13 (Lambda)             (None, 12, 12, 1088) 0           block17_12_ac[0][0]              \n",
            "                                                                 block17_13_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_13_ac (Activation)      (None, 12, 12, 1088) 0           block17_13[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_571 (Conv2D)             (None, 12, 12, 128)  139264      block17_13_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_537 (BatchN (None, 12, 12, 128)  384         conv2d_571[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1055 (Activation)    (None, 12, 12, 128)  0           batch_normalization_537[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_572 (Conv2D)             (None, 12, 12, 160)  143360      activation_1055[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_538 (BatchN (None, 12, 12, 160)  480         conv2d_572[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1056 (Activation)    (None, 12, 12, 160)  0           batch_normalization_538[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_570 (Conv2D)             (None, 12, 12, 192)  208896      block17_13_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_573 (Conv2D)             (None, 12, 12, 192)  215040      activation_1056[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_536 (BatchN (None, 12, 12, 192)  576         conv2d_570[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_539 (BatchN (None, 12, 12, 192)  576         conv2d_573[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1054 (Activation)    (None, 12, 12, 192)  0           batch_normalization_536[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1057 (Activation)    (None, 12, 12, 192)  0           batch_normalization_539[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_14_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1054[0][0]            \n",
            "                                                                 activation_1057[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_14_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_14_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_14 (Lambda)             (None, 12, 12, 1088) 0           block17_13_ac[0][0]              \n",
            "                                                                 block17_14_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_14_ac (Activation)      (None, 12, 12, 1088) 0           block17_14[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_575 (Conv2D)             (None, 12, 12, 128)  139264      block17_14_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_541 (BatchN (None, 12, 12, 128)  384         conv2d_575[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1059 (Activation)    (None, 12, 12, 128)  0           batch_normalization_541[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_576 (Conv2D)             (None, 12, 12, 160)  143360      activation_1059[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_542 (BatchN (None, 12, 12, 160)  480         conv2d_576[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1060 (Activation)    (None, 12, 12, 160)  0           batch_normalization_542[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_574 (Conv2D)             (None, 12, 12, 192)  208896      block17_14_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_577 (Conv2D)             (None, 12, 12, 192)  215040      activation_1060[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_540 (BatchN (None, 12, 12, 192)  576         conv2d_574[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_543 (BatchN (None, 12, 12, 192)  576         conv2d_577[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1058 (Activation)    (None, 12, 12, 192)  0           batch_normalization_540[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1061 (Activation)    (None, 12, 12, 192)  0           batch_normalization_543[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_15_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1058[0][0]            \n",
            "                                                                 activation_1061[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_15_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_15_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_15 (Lambda)             (None, 12, 12, 1088) 0           block17_14_ac[0][0]              \n",
            "                                                                 block17_15_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_15_ac (Activation)      (None, 12, 12, 1088) 0           block17_15[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_579 (Conv2D)             (None, 12, 12, 128)  139264      block17_15_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_545 (BatchN (None, 12, 12, 128)  384         conv2d_579[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1063 (Activation)    (None, 12, 12, 128)  0           batch_normalization_545[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_580 (Conv2D)             (None, 12, 12, 160)  143360      activation_1063[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_546 (BatchN (None, 12, 12, 160)  480         conv2d_580[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1064 (Activation)    (None, 12, 12, 160)  0           batch_normalization_546[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_578 (Conv2D)             (None, 12, 12, 192)  208896      block17_15_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_581 (Conv2D)             (None, 12, 12, 192)  215040      activation_1064[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_544 (BatchN (None, 12, 12, 192)  576         conv2d_578[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_547 (BatchN (None, 12, 12, 192)  576         conv2d_581[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1062 (Activation)    (None, 12, 12, 192)  0           batch_normalization_544[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1065 (Activation)    (None, 12, 12, 192)  0           batch_normalization_547[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_16_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1062[0][0]            \n",
            "                                                                 activation_1065[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_16_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_16_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_16 (Lambda)             (None, 12, 12, 1088) 0           block17_15_ac[0][0]              \n",
            "                                                                 block17_16_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_16_ac (Activation)      (None, 12, 12, 1088) 0           block17_16[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_583 (Conv2D)             (None, 12, 12, 128)  139264      block17_16_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_549 (BatchN (None, 12, 12, 128)  384         conv2d_583[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1067 (Activation)    (None, 12, 12, 128)  0           batch_normalization_549[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_584 (Conv2D)             (None, 12, 12, 160)  143360      activation_1067[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_550 (BatchN (None, 12, 12, 160)  480         conv2d_584[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1068 (Activation)    (None, 12, 12, 160)  0           batch_normalization_550[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_582 (Conv2D)             (None, 12, 12, 192)  208896      block17_16_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_585 (Conv2D)             (None, 12, 12, 192)  215040      activation_1068[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_548 (BatchN (None, 12, 12, 192)  576         conv2d_582[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_551 (BatchN (None, 12, 12, 192)  576         conv2d_585[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1066 (Activation)    (None, 12, 12, 192)  0           batch_normalization_548[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1069 (Activation)    (None, 12, 12, 192)  0           batch_normalization_551[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_17_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1066[0][0]            \n",
            "                                                                 activation_1069[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_17_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_17_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_17 (Lambda)             (None, 12, 12, 1088) 0           block17_16_ac[0][0]              \n",
            "                                                                 block17_17_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_17_ac (Activation)      (None, 12, 12, 1088) 0           block17_17[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_587 (Conv2D)             (None, 12, 12, 128)  139264      block17_17_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_553 (BatchN (None, 12, 12, 128)  384         conv2d_587[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1071 (Activation)    (None, 12, 12, 128)  0           batch_normalization_553[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_588 (Conv2D)             (None, 12, 12, 160)  143360      activation_1071[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_554 (BatchN (None, 12, 12, 160)  480         conv2d_588[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1072 (Activation)    (None, 12, 12, 160)  0           batch_normalization_554[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_586 (Conv2D)             (None, 12, 12, 192)  208896      block17_17_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_589 (Conv2D)             (None, 12, 12, 192)  215040      activation_1072[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_552 (BatchN (None, 12, 12, 192)  576         conv2d_586[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_555 (BatchN (None, 12, 12, 192)  576         conv2d_589[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1070 (Activation)    (None, 12, 12, 192)  0           batch_normalization_552[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1073 (Activation)    (None, 12, 12, 192)  0           batch_normalization_555[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_18_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1070[0][0]            \n",
            "                                                                 activation_1073[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_18_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_18_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_18 (Lambda)             (None, 12, 12, 1088) 0           block17_17_ac[0][0]              \n",
            "                                                                 block17_18_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_18_ac (Activation)      (None, 12, 12, 1088) 0           block17_18[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_591 (Conv2D)             (None, 12, 12, 128)  139264      block17_18_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_557 (BatchN (None, 12, 12, 128)  384         conv2d_591[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1075 (Activation)    (None, 12, 12, 128)  0           batch_normalization_557[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_592 (Conv2D)             (None, 12, 12, 160)  143360      activation_1075[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_558 (BatchN (None, 12, 12, 160)  480         conv2d_592[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1076 (Activation)    (None, 12, 12, 160)  0           batch_normalization_558[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_590 (Conv2D)             (None, 12, 12, 192)  208896      block17_18_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_593 (Conv2D)             (None, 12, 12, 192)  215040      activation_1076[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_556 (BatchN (None, 12, 12, 192)  576         conv2d_590[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_559 (BatchN (None, 12, 12, 192)  576         conv2d_593[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1074 (Activation)    (None, 12, 12, 192)  0           batch_normalization_556[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1077 (Activation)    (None, 12, 12, 192)  0           batch_normalization_559[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_19_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1074[0][0]            \n",
            "                                                                 activation_1077[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_19_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_19_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_19 (Lambda)             (None, 12, 12, 1088) 0           block17_18_ac[0][0]              \n",
            "                                                                 block17_19_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_19_ac (Activation)      (None, 12, 12, 1088) 0           block17_19[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_595 (Conv2D)             (None, 12, 12, 128)  139264      block17_19_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_561 (BatchN (None, 12, 12, 128)  384         conv2d_595[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1079 (Activation)    (None, 12, 12, 128)  0           batch_normalization_561[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_596 (Conv2D)             (None, 12, 12, 160)  143360      activation_1079[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_562 (BatchN (None, 12, 12, 160)  480         conv2d_596[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1080 (Activation)    (None, 12, 12, 160)  0           batch_normalization_562[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_594 (Conv2D)             (None, 12, 12, 192)  208896      block17_19_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_597 (Conv2D)             (None, 12, 12, 192)  215040      activation_1080[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_560 (BatchN (None, 12, 12, 192)  576         conv2d_594[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_563 (BatchN (None, 12, 12, 192)  576         conv2d_597[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1078 (Activation)    (None, 12, 12, 192)  0           batch_normalization_560[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1081 (Activation)    (None, 12, 12, 192)  0           batch_normalization_563[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block17_20_mixed (Concatenate)  (None, 12, 12, 384)  0           activation_1078[0][0]            \n",
            "                                                                 activation_1081[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_20_conv (Conv2D)        (None, 12, 12, 1088) 418880      block17_20_mixed[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "block17_20 (Lambda)             (None, 12, 12, 1088) 0           block17_19_ac[0][0]              \n",
            "                                                                 block17_20_conv[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block17_20_ac (Activation)      (None, 12, 12, 1088) 0           block17_20[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_602 (Conv2D)             (None, 12, 12, 256)  278528      block17_20_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_568 (BatchN (None, 12, 12, 256)  768         conv2d_602[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1086 (Activation)    (None, 12, 12, 256)  0           batch_normalization_568[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_598 (Conv2D)             (None, 12, 12, 256)  278528      block17_20_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_600 (Conv2D)             (None, 12, 12, 256)  278528      block17_20_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_603 (Conv2D)             (None, 12, 12, 288)  663552      activation_1086[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_564 (BatchN (None, 12, 12, 256)  768         conv2d_598[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_566 (BatchN (None, 12, 12, 256)  768         conv2d_600[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_569 (BatchN (None, 12, 12, 288)  864         conv2d_603[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1082 (Activation)    (None, 12, 12, 256)  0           batch_normalization_564[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1084 (Activation)    (None, 12, 12, 256)  0           batch_normalization_566[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1087 (Activation)    (None, 12, 12, 288)  0           batch_normalization_569[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_599 (Conv2D)             (None, 5, 5, 384)    884736      activation_1082[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_601 (Conv2D)             (None, 5, 5, 288)    663552      activation_1084[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_604 (Conv2D)             (None, 5, 5, 320)    829440      activation_1087[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_565 (BatchN (None, 5, 5, 384)    1152        conv2d_599[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_567 (BatchN (None, 5, 5, 288)    864         conv2d_601[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_570 (BatchN (None, 5, 5, 320)    960         conv2d_604[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1083 (Activation)    (None, 5, 5, 384)    0           batch_normalization_565[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1085 (Activation)    (None, 5, 5, 288)    0           batch_normalization_567[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1088 (Activation)    (None, 5, 5, 320)    0           batch_normalization_570[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_13 (MaxPooling2D) (None, 5, 5, 1088)   0           block17_20_ac[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "mixed_7a (Concatenate)          (None, 5, 5, 2080)   0           activation_1083[0][0]            \n",
            "                                                                 activation_1085[0][0]            \n",
            "                                                                 activation_1088[0][0]            \n",
            "                                                                 max_pooling2d_13[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_606 (Conv2D)             (None, 5, 5, 192)    399360      mixed_7a[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_572 (BatchN (None, 5, 5, 192)    576         conv2d_606[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1090 (Activation)    (None, 5, 5, 192)    0           batch_normalization_572[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_607 (Conv2D)             (None, 5, 5, 224)    129024      activation_1090[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_573 (BatchN (None, 5, 5, 224)    672         conv2d_607[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1091 (Activation)    (None, 5, 5, 224)    0           batch_normalization_573[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_605 (Conv2D)             (None, 5, 5, 192)    399360      mixed_7a[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_608 (Conv2D)             (None, 5, 5, 256)    172032      activation_1091[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_571 (BatchN (None, 5, 5, 192)    576         conv2d_605[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_574 (BatchN (None, 5, 5, 256)    768         conv2d_608[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1089 (Activation)    (None, 5, 5, 192)    0           batch_normalization_571[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1092 (Activation)    (None, 5, 5, 256)    0           batch_normalization_574[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_1_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1089[0][0]            \n",
            "                                                                 activation_1092[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_1_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_1_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_1 (Lambda)               (None, 5, 5, 2080)   0           mixed_7a[0][0]                   \n",
            "                                                                 block8_1_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_1_ac (Activation)        (None, 5, 5, 2080)   0           block8_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_610 (Conv2D)             (None, 5, 5, 192)    399360      block8_1_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_576 (BatchN (None, 5, 5, 192)    576         conv2d_610[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1094 (Activation)    (None, 5, 5, 192)    0           batch_normalization_576[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_611 (Conv2D)             (None, 5, 5, 224)    129024      activation_1094[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_577 (BatchN (None, 5, 5, 224)    672         conv2d_611[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1095 (Activation)    (None, 5, 5, 224)    0           batch_normalization_577[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_609 (Conv2D)             (None, 5, 5, 192)    399360      block8_1_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_612 (Conv2D)             (None, 5, 5, 256)    172032      activation_1095[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_575 (BatchN (None, 5, 5, 192)    576         conv2d_609[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_578 (BatchN (None, 5, 5, 256)    768         conv2d_612[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1093 (Activation)    (None, 5, 5, 192)    0           batch_normalization_575[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1096 (Activation)    (None, 5, 5, 256)    0           batch_normalization_578[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_2_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1093[0][0]            \n",
            "                                                                 activation_1096[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_2_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_2_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_2 (Lambda)               (None, 5, 5, 2080)   0           block8_1_ac[0][0]                \n",
            "                                                                 block8_2_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_2_ac (Activation)        (None, 5, 5, 2080)   0           block8_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_614 (Conv2D)             (None, 5, 5, 192)    399360      block8_2_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_580 (BatchN (None, 5, 5, 192)    576         conv2d_614[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1098 (Activation)    (None, 5, 5, 192)    0           batch_normalization_580[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_615 (Conv2D)             (None, 5, 5, 224)    129024      activation_1098[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_581 (BatchN (None, 5, 5, 224)    672         conv2d_615[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1099 (Activation)    (None, 5, 5, 224)    0           batch_normalization_581[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_613 (Conv2D)             (None, 5, 5, 192)    399360      block8_2_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_616 (Conv2D)             (None, 5, 5, 256)    172032      activation_1099[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_579 (BatchN (None, 5, 5, 192)    576         conv2d_613[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_582 (BatchN (None, 5, 5, 256)    768         conv2d_616[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1097 (Activation)    (None, 5, 5, 192)    0           batch_normalization_579[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1100 (Activation)    (None, 5, 5, 256)    0           batch_normalization_582[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_3_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1097[0][0]            \n",
            "                                                                 activation_1100[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_3_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_3_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_3 (Lambda)               (None, 5, 5, 2080)   0           block8_2_ac[0][0]                \n",
            "                                                                 block8_3_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_3_ac (Activation)        (None, 5, 5, 2080)   0           block8_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_618 (Conv2D)             (None, 5, 5, 192)    399360      block8_3_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_584 (BatchN (None, 5, 5, 192)    576         conv2d_618[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1102 (Activation)    (None, 5, 5, 192)    0           batch_normalization_584[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_619 (Conv2D)             (None, 5, 5, 224)    129024      activation_1102[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_585 (BatchN (None, 5, 5, 224)    672         conv2d_619[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1103 (Activation)    (None, 5, 5, 224)    0           batch_normalization_585[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_617 (Conv2D)             (None, 5, 5, 192)    399360      block8_3_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_620 (Conv2D)             (None, 5, 5, 256)    172032      activation_1103[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_583 (BatchN (None, 5, 5, 192)    576         conv2d_617[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_586 (BatchN (None, 5, 5, 256)    768         conv2d_620[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1101 (Activation)    (None, 5, 5, 192)    0           batch_normalization_583[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1104 (Activation)    (None, 5, 5, 256)    0           batch_normalization_586[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_4_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1101[0][0]            \n",
            "                                                                 activation_1104[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_4_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_4_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_4 (Lambda)               (None, 5, 5, 2080)   0           block8_3_ac[0][0]                \n",
            "                                                                 block8_4_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_4_ac (Activation)        (None, 5, 5, 2080)   0           block8_4[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_622 (Conv2D)             (None, 5, 5, 192)    399360      block8_4_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_588 (BatchN (None, 5, 5, 192)    576         conv2d_622[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1106 (Activation)    (None, 5, 5, 192)    0           batch_normalization_588[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_623 (Conv2D)             (None, 5, 5, 224)    129024      activation_1106[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_589 (BatchN (None, 5, 5, 224)    672         conv2d_623[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1107 (Activation)    (None, 5, 5, 224)    0           batch_normalization_589[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_621 (Conv2D)             (None, 5, 5, 192)    399360      block8_4_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_624 (Conv2D)             (None, 5, 5, 256)    172032      activation_1107[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_587 (BatchN (None, 5, 5, 192)    576         conv2d_621[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_590 (BatchN (None, 5, 5, 256)    768         conv2d_624[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1105 (Activation)    (None, 5, 5, 192)    0           batch_normalization_587[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1108 (Activation)    (None, 5, 5, 256)    0           batch_normalization_590[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_5_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1105[0][0]            \n",
            "                                                                 activation_1108[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_5_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_5_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_5 (Lambda)               (None, 5, 5, 2080)   0           block8_4_ac[0][0]                \n",
            "                                                                 block8_5_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_5_ac (Activation)        (None, 5, 5, 2080)   0           block8_5[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_626 (Conv2D)             (None, 5, 5, 192)    399360      block8_5_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_592 (BatchN (None, 5, 5, 192)    576         conv2d_626[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1110 (Activation)    (None, 5, 5, 192)    0           batch_normalization_592[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_627 (Conv2D)             (None, 5, 5, 224)    129024      activation_1110[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_593 (BatchN (None, 5, 5, 224)    672         conv2d_627[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1111 (Activation)    (None, 5, 5, 224)    0           batch_normalization_593[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_625 (Conv2D)             (None, 5, 5, 192)    399360      block8_5_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_628 (Conv2D)             (None, 5, 5, 256)    172032      activation_1111[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_591 (BatchN (None, 5, 5, 192)    576         conv2d_625[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_594 (BatchN (None, 5, 5, 256)    768         conv2d_628[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1109 (Activation)    (None, 5, 5, 192)    0           batch_normalization_591[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1112 (Activation)    (None, 5, 5, 256)    0           batch_normalization_594[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_6_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1109[0][0]            \n",
            "                                                                 activation_1112[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_6_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_6_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_6 (Lambda)               (None, 5, 5, 2080)   0           block8_5_ac[0][0]                \n",
            "                                                                 block8_6_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_6_ac (Activation)        (None, 5, 5, 2080)   0           block8_6[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_630 (Conv2D)             (None, 5, 5, 192)    399360      block8_6_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_596 (BatchN (None, 5, 5, 192)    576         conv2d_630[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1114 (Activation)    (None, 5, 5, 192)    0           batch_normalization_596[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_631 (Conv2D)             (None, 5, 5, 224)    129024      activation_1114[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_597 (BatchN (None, 5, 5, 224)    672         conv2d_631[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1115 (Activation)    (None, 5, 5, 224)    0           batch_normalization_597[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_629 (Conv2D)             (None, 5, 5, 192)    399360      block8_6_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_632 (Conv2D)             (None, 5, 5, 256)    172032      activation_1115[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_595 (BatchN (None, 5, 5, 192)    576         conv2d_629[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_598 (BatchN (None, 5, 5, 256)    768         conv2d_632[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1113 (Activation)    (None, 5, 5, 192)    0           batch_normalization_595[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1116 (Activation)    (None, 5, 5, 256)    0           batch_normalization_598[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_7_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1113[0][0]            \n",
            "                                                                 activation_1116[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_7_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_7_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_7 (Lambda)               (None, 5, 5, 2080)   0           block8_6_ac[0][0]                \n",
            "                                                                 block8_7_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_7_ac (Activation)        (None, 5, 5, 2080)   0           block8_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_634 (Conv2D)             (None, 5, 5, 192)    399360      block8_7_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_600 (BatchN (None, 5, 5, 192)    576         conv2d_634[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1118 (Activation)    (None, 5, 5, 192)    0           batch_normalization_600[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_635 (Conv2D)             (None, 5, 5, 224)    129024      activation_1118[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_601 (BatchN (None, 5, 5, 224)    672         conv2d_635[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1119 (Activation)    (None, 5, 5, 224)    0           batch_normalization_601[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_633 (Conv2D)             (None, 5, 5, 192)    399360      block8_7_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_636 (Conv2D)             (None, 5, 5, 256)    172032      activation_1119[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_599 (BatchN (None, 5, 5, 192)    576         conv2d_633[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_602 (BatchN (None, 5, 5, 256)    768         conv2d_636[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1117 (Activation)    (None, 5, 5, 192)    0           batch_normalization_599[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1120 (Activation)    (None, 5, 5, 256)    0           batch_normalization_602[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_8_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1117[0][0]            \n",
            "                                                                 activation_1120[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_8_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_8_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_8 (Lambda)               (None, 5, 5, 2080)   0           block8_7_ac[0][0]                \n",
            "                                                                 block8_8_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_8_ac (Activation)        (None, 5, 5, 2080)   0           block8_8[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_638 (Conv2D)             (None, 5, 5, 192)    399360      block8_8_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_604 (BatchN (None, 5, 5, 192)    576         conv2d_638[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1122 (Activation)    (None, 5, 5, 192)    0           batch_normalization_604[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_639 (Conv2D)             (None, 5, 5, 224)    129024      activation_1122[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_605 (BatchN (None, 5, 5, 224)    672         conv2d_639[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1123 (Activation)    (None, 5, 5, 224)    0           batch_normalization_605[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_637 (Conv2D)             (None, 5, 5, 192)    399360      block8_8_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_640 (Conv2D)             (None, 5, 5, 256)    172032      activation_1123[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_603 (BatchN (None, 5, 5, 192)    576         conv2d_637[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_606 (BatchN (None, 5, 5, 256)    768         conv2d_640[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1121 (Activation)    (None, 5, 5, 192)    0           batch_normalization_603[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1124 (Activation)    (None, 5, 5, 256)    0           batch_normalization_606[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_9_mixed (Concatenate)    (None, 5, 5, 448)    0           activation_1121[0][0]            \n",
            "                                                                 activation_1124[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_9_conv (Conv2D)          (None, 5, 5, 2080)   933920      block8_9_mixed[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "block8_9 (Lambda)               (None, 5, 5, 2080)   0           block8_8_ac[0][0]                \n",
            "                                                                 block8_9_conv[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "block8_9_ac (Activation)        (None, 5, 5, 2080)   0           block8_9[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_642 (Conv2D)             (None, 5, 5, 192)    399360      block8_9_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_608 (BatchN (None, 5, 5, 192)    576         conv2d_642[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1126 (Activation)    (None, 5, 5, 192)    0           batch_normalization_608[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_643 (Conv2D)             (None, 5, 5, 224)    129024      activation_1126[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_609 (BatchN (None, 5, 5, 224)    672         conv2d_643[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1127 (Activation)    (None, 5, 5, 224)    0           batch_normalization_609[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_641 (Conv2D)             (None, 5, 5, 192)    399360      block8_9_ac[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_644 (Conv2D)             (None, 5, 5, 256)    172032      activation_1127[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_607 (BatchN (None, 5, 5, 192)    576         conv2d_641[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_610 (BatchN (None, 5, 5, 256)    768         conv2d_644[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "activation_1125 (Activation)    (None, 5, 5, 192)    0           batch_normalization_607[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "activation_1128 (Activation)    (None, 5, 5, 256)    0           batch_normalization_610[0][0]    \n",
            "__________________________________________________________________________________________________\n",
            "block8_10_mixed (Concatenate)   (None, 5, 5, 448)    0           activation_1125[0][0]            \n",
            "                                                                 activation_1128[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_10_conv (Conv2D)         (None, 5, 5, 2080)   933920      block8_10_mixed[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "block8_10 (Lambda)              (None, 5, 5, 2080)   0           block8_9_ac[0][0]                \n",
            "                                                                 block8_10_conv[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv_7b (Conv2D)                (None, 5, 5, 1536)   3194880     block8_10[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv_7b_bn (BatchNormalization) (None, 5, 5, 1536)   4608        conv_7b[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "conv_7b_ac (Activation)         (None, 5, 5, 1536)   0           conv_7b_bn[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d_13 (AveragePo (None, 1, 1, 1536)   0           conv_7b_ac[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "flatten (Flatten)               (None, 1536)         0           average_pooling2d_13[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "dense_16 (Dense)                (None, 64)           98368       flatten[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout_10 (Dropout)            (None, 64)           0           dense_16[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "dense_17 (Dense)                (None, 2)            130         dropout_10[0][0]                 \n",
            "==================================================================================================\n",
            "Total params: 54,435,234\n",
            "Trainable params: 98,498\n",
            "Non-trainable params: 54,336,736\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O5tmykjuUI99",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(loss=\"categorical_crossentropy\", \n",
        "              optimizer=Nadam(),\n",
        "              metrics=[\"accuracy\"])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "knh9uxF_ULmG",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "f8abdea8-a343-4ad9-eabf-d6d56b314d19"
      },
      "source": [
        "H4 = model.fit(train_augmented_data, \n",
        "                        steps_per_epoch=8,\n",
        "                        validation_data=validation_data,\n",
        "                        validation_steps=8,\n",
        "                        epochs=50)"
      ],
      "execution_count": 84,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "8/8 [==============================] - 7s 814ms/step - loss: 0.8678 - accuracy: 0.6398 - val_loss: 0.1143 - val_accuracy: 0.9750\n",
            "Epoch 2/50\n",
            "8/8 [==============================] - 4s 471ms/step - loss: 0.3844 - accuracy: 0.8390 - val_loss: 0.0746 - val_accuracy: 0.9750\n",
            "Epoch 3/50\n",
            "8/8 [==============================] - 4s 471ms/step - loss: 0.2324 - accuracy: 0.8941 - val_loss: 0.0540 - val_accuracy: 1.0000\n",
            "Epoch 4/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.1571 - accuracy: 0.9449 - val_loss: 0.0523 - val_accuracy: 0.9875\n",
            "Epoch 5/50\n",
            "8/8 [==============================] - 4s 468ms/step - loss: 0.1490 - accuracy: 0.9364 - val_loss: 0.0442 - val_accuracy: 0.9875\n",
            "Epoch 6/50\n",
            "8/8 [==============================] - 4s 465ms/step - loss: 0.1998 - accuracy: 0.9068 - val_loss: 0.0264 - val_accuracy: 1.0000\n",
            "Epoch 7/50\n",
            "8/8 [==============================] - 4s 509ms/step - loss: 0.1416 - accuracy: 0.9364 - val_loss: 0.0244 - val_accuracy: 1.0000\n",
            "Epoch 8/50\n",
            "8/8 [==============================] - 4s 467ms/step - loss: 0.1360 - accuracy: 0.9449 - val_loss: 0.0284 - val_accuracy: 1.0000\n",
            "Epoch 9/50\n",
            "8/8 [==============================] - 4s 465ms/step - loss: 0.1155 - accuracy: 0.9364 - val_loss: 0.0310 - val_accuracy: 0.9875\n",
            "Epoch 10/50\n",
            "8/8 [==============================] - 4s 474ms/step - loss: 0.1691 - accuracy: 0.9407 - val_loss: 0.0277 - val_accuracy: 1.0000\n",
            "Epoch 11/50\n",
            "8/8 [==============================] - 4s 457ms/step - loss: 0.1233 - accuracy: 0.9661 - val_loss: 0.0224 - val_accuracy: 1.0000\n",
            "Epoch 12/50\n",
            "8/8 [==============================] - 4s 508ms/step - loss: 0.1496 - accuracy: 0.9449 - val_loss: 0.0319 - val_accuracy: 0.9875\n",
            "Epoch 13/50\n",
            "8/8 [==============================] - 4s 467ms/step - loss: 0.0890 - accuracy: 0.9746 - val_loss: 0.0185 - val_accuracy: 1.0000\n",
            "Epoch 14/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.0990 - accuracy: 0.9576 - val_loss: 0.0184 - val_accuracy: 1.0000\n",
            "Epoch 15/50\n",
            "8/8 [==============================] - 4s 467ms/step - loss: 0.1404 - accuracy: 0.9534 - val_loss: 0.0217 - val_accuracy: 1.0000\n",
            "Epoch 16/50\n",
            "8/8 [==============================] - 4s 511ms/step - loss: 0.1052 - accuracy: 0.9619 - val_loss: 0.0170 - val_accuracy: 1.0000\n",
            "Epoch 17/50\n",
            "8/8 [==============================] - 4s 514ms/step - loss: 0.1301 - accuracy: 0.9534 - val_loss: 0.0185 - val_accuracy: 1.0000\n",
            "Epoch 18/50\n",
            "8/8 [==============================] - 4s 473ms/step - loss: 0.1248 - accuracy: 0.9619 - val_loss: 0.0257 - val_accuracy: 1.0000\n",
            "Epoch 19/50\n",
            "8/8 [==============================] - 4s 471ms/step - loss: 0.1097 - accuracy: 0.9661 - val_loss: 0.0200 - val_accuracy: 1.0000\n",
            "Epoch 20/50\n",
            "8/8 [==============================] - 4s 476ms/step - loss: 0.1037 - accuracy: 0.9619 - val_loss: 0.0139 - val_accuracy: 1.0000\n",
            "Epoch 21/50\n",
            "8/8 [==============================] - 4s 469ms/step - loss: 0.0846 - accuracy: 0.9661 - val_loss: 0.0270 - val_accuracy: 0.9875\n",
            "Epoch 22/50\n",
            "8/8 [==============================] - 4s 456ms/step - loss: 0.0715 - accuracy: 0.9788 - val_loss: 0.0371 - val_accuracy: 0.9750\n",
            "Epoch 23/50\n",
            "8/8 [==============================] - 4s 466ms/step - loss: 0.0681 - accuracy: 0.9746 - val_loss: 0.0159 - val_accuracy: 1.0000\n",
            "Epoch 24/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.1461 - accuracy: 0.9576 - val_loss: 0.0147 - val_accuracy: 1.0000\n",
            "Epoch 25/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.0947 - accuracy: 0.9746 - val_loss: 0.0123 - val_accuracy: 1.0000\n",
            "Epoch 26/50\n",
            "8/8 [==============================] - 4s 463ms/step - loss: 0.1197 - accuracy: 0.9534 - val_loss: 0.0145 - val_accuracy: 1.0000\n",
            "Epoch 27/50\n",
            "8/8 [==============================] - 4s 463ms/step - loss: 0.0761 - accuracy: 0.9746 - val_loss: 0.0145 - val_accuracy: 1.0000\n",
            "Epoch 28/50\n",
            "8/8 [==============================] - 4s 510ms/step - loss: 0.0535 - accuracy: 0.9788 - val_loss: 0.0106 - val_accuracy: 1.0000\n",
            "Epoch 29/50\n",
            "8/8 [==============================] - 4s 508ms/step - loss: 0.1156 - accuracy: 0.9534 - val_loss: 0.0128 - val_accuracy: 1.0000\n",
            "Epoch 30/50\n",
            "8/8 [==============================] - 4s 467ms/step - loss: 0.1362 - accuracy: 0.9576 - val_loss: 0.0169 - val_accuracy: 1.0000\n",
            "Epoch 31/50\n",
            "8/8 [==============================] - 4s 469ms/step - loss: 0.0978 - accuracy: 0.9619 - val_loss: 0.0155 - val_accuracy: 1.0000\n",
            "Epoch 32/50\n",
            "8/8 [==============================] - 4s 509ms/step - loss: 0.0835 - accuracy: 0.9661 - val_loss: 0.0125 - val_accuracy: 1.0000\n",
            "Epoch 33/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.1007 - accuracy: 0.9661 - val_loss: 0.0098 - val_accuracy: 1.0000\n",
            "Epoch 34/50\n",
            "8/8 [==============================] - 4s 466ms/step - loss: 0.0974 - accuracy: 0.9746 - val_loss: 0.0125 - val_accuracy: 1.0000\n",
            "Epoch 35/50\n",
            "8/8 [==============================] - 4s 469ms/step - loss: 0.0717 - accuracy: 0.9746 - val_loss: 0.0087 - val_accuracy: 1.0000\n",
            "Epoch 36/50\n",
            "8/8 [==============================] - 4s 470ms/step - loss: 0.0982 - accuracy: 0.9746 - val_loss: 0.0117 - val_accuracy: 1.0000\n",
            "Epoch 37/50\n",
            "8/8 [==============================] - 4s 519ms/step - loss: 0.0709 - accuracy: 0.9788 - val_loss: 0.0092 - val_accuracy: 1.0000\n",
            "Epoch 38/50\n",
            "8/8 [==============================] - 4s 514ms/step - loss: 0.0967 - accuracy: 0.9788 - val_loss: 0.0298 - val_accuracy: 0.9875\n",
            "Epoch 39/50\n",
            "8/8 [==============================] - 4s 513ms/step - loss: 0.0976 - accuracy: 0.9661 - val_loss: 0.0118 - val_accuracy: 1.0000\n",
            "Epoch 40/50\n",
            "8/8 [==============================] - 4s 465ms/step - loss: 0.0525 - accuracy: 0.9831 - val_loss: 0.0077 - val_accuracy: 1.0000\n",
            "Epoch 41/50\n",
            "8/8 [==============================] - 4s 507ms/step - loss: 0.0660 - accuracy: 0.9746 - val_loss: 0.0066 - val_accuracy: 1.0000\n",
            "Epoch 42/50\n",
            "8/8 [==============================] - 4s 467ms/step - loss: 0.0970 - accuracy: 0.9703 - val_loss: 0.0131 - val_accuracy: 1.0000\n",
            "Epoch 43/50\n",
            "8/8 [==============================] - 4s 468ms/step - loss: 0.1074 - accuracy: 0.9661 - val_loss: 0.0142 - val_accuracy: 1.0000\n",
            "Epoch 44/50\n",
            "8/8 [==============================] - 4s 468ms/step - loss: 0.0798 - accuracy: 0.9703 - val_loss: 0.0111 - val_accuracy: 1.0000\n",
            "Epoch 45/50\n",
            "8/8 [==============================] - 4s 471ms/step - loss: 0.0986 - accuracy: 0.9746 - val_loss: 0.0118 - val_accuracy: 1.0000\n",
            "Epoch 46/50\n",
            "8/8 [==============================] - 4s 476ms/step - loss: 0.0856 - accuracy: 0.9746 - val_loss: 0.0194 - val_accuracy: 1.0000\n",
            "Epoch 47/50\n",
            "8/8 [==============================] - 4s 474ms/step - loss: 0.0746 - accuracy: 0.9746 - val_loss: 0.0135 - val_accuracy: 1.0000\n",
            "Epoch 48/50\n",
            "8/8 [==============================] - 4s 466ms/step - loss: 0.1111 - accuracy: 0.9661 - val_loss: 0.0117 - val_accuracy: 1.0000\n",
            "Epoch 49/50\n",
            "8/8 [==============================] - 4s 467ms/step - loss: 0.0935 - accuracy: 0.9703 - val_loss: 0.0117 - val_accuracy: 1.0000\n",
            "Epoch 50/50\n",
            "8/8 [==============================] - 4s 464ms/step - loss: 0.0988 - accuracy: 0.9619 - val_loss: 0.0190 - val_accuracy: 1.0000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Lf3Ec8zdTVxa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "outputId": "a913fe93-6a48-44c2-ce5d-7dc6ca358aa4"
      },
      "source": [
        "plt.plot(H4.history['accuracy'], label='accuracy')\n",
        "plt.plot(H4.history['val_accuracy'], label = 'val_accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('loss')\n",
        "plt.ylim([0., 1])\n",
        "plt.legend(loc='upper right')"
      ],
      "execution_count": 85,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f992ca4fd30>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 85
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GPGJtcMBTVNS",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "outputId": "9bdb79e4-44ae-4aee-ae39-d32da1484bd1"
      },
      "source": [
        "plt.plot(H4.history['loss'], label='loss')\n",
        "plt.plot(H4.history['val_loss'], label = 'val_loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('loss')\n",
        "plt.ylim([0., 1])\n",
        "plt.legend(loc='upper right')"
      ],
      "execution_count": 86,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f992ca32e10>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 86
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yezapVs0WEpx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "bf559d82-4c3a-4931-e070-3a2f7d4831c8"
      },
      "source": [
        "predictions = model.predict(test_data,\n",
        "                     batch_size=10,\n",
        "                     verbose=1)\n",
        "y_pred = np.argmax(predictions, axis=1)\n",
        "print(confusion_matrix(test_data.classes, y_pred))\n",
        "print(classification_report(test_data.classes, y_pred, target_names = ['covid', 'healthy']))"
      ],
      "execution_count": 87,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "8/8 [==============================] - 1s 87ms/step\n",
            "[[40  0]\n",
            " [ 0 40]]\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "       covid       1.00      1.00      1.00        40\n",
            "     healthy       1.00      1.00      1.00        40\n",
            "\n",
            "    accuracy                           1.00        80\n",
            "   macro avg       1.00      1.00      1.00        80\n",
            "weighted avg       1.00      1.00      1.00        80\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3Ea3MAOfLW-B",
        "colab_type": "text"
      },
      "source": [
        "Como podemos observar, con este último modelo hemos conseguido obtener el mejor resultado para el conjunto de test, llegando al 100% de acierto de clasificación."
      ]
    }
  ]
}