Model_01-DAG.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from os.path import join as opj\n",
    "from matplotlib import pyplot as plt\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "import pylab\n",
    "plt.rcParams['figure.figsize'] = 10, 10\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Load the data.\n",
    "train = pd.read_json(\"input/train.json\")\n",
    "target_train = np.array(train['is_iceberg'])\n",
    "test = pd.read_json(\"input/test.json\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Generate the training data\n",
    "#Create 3 bands having HH, HV and avg of both\n",
    "X_band_1=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train[\"band_1\"]])\n",
    "X_band_2=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train[\"band_2\"]])\n",
    "X_train = np.concatenate([X_band_1[:, :, :, np.newaxis], X_band_2[:, :, :, np.newaxis],((X_band_1+X_band_2)/2)[:, :, :, np.newaxis]], axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_band_test_1=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in test[\"band_1\"]])\n",
    "X_band_test_2=np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in test[\"band_2\"]])\n",
    "X_band_test_3=(X_band_test_1+X_band_test_2)/2\n",
    "#X_band_test_3=np.array([np.full((75, 75), angel).astype(np.float32) for angel in test[\"inc_angle\"]])\n",
    "X_test = np.concatenate([X_band_test_1[:, :, :, np.newaxis]\n",
    "                          , X_band_test_2[:, :, :, np.newaxis]\n",
    "                         , X_band_test_3[:, :, :, np.newaxis]], axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.inc_angle = train.inc_angle.replace('na', 0)\n",
    "idx_tr = np.where(train.inc_angle > 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# target_train = target_train[idx_tr[0]]\n",
    "# X_train = X_train[idx_tr[0],...]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_angle_train = np.array(train.inc_angle)\n",
    "X_angle_test = np.array(test.inc_angle)\n",
    "y_train = np.array(train[\"is_iceberg\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a train and validation split, 75% of data used in training\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_valid, X_angle_train, X_angle_valid, y_train, y_valid = train_test_split(X_train,\n",
    "                                    X_angle_train, y_train, random_state=1, train_size=0.75)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "#Import Keras.\n",
    "from matplotlib import pyplot\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D, Dense, Dropout, Input, Flatten, Activation\n",
    "from keras.layers import GlobalMaxPooling2D\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.merge import Concatenate\n",
    "from keras.models import Model\n",
    "from keras import initializers\n",
    "from keras.optimizers import Adam\n",
    "from keras.callbacks import ModelCheckpoint, Callback, EarlyStopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# from keras.models import Model\n",
    "# from keras.layers import Input, Dense, Reshape, concatenate, Conv2D, Flatten, MaxPooling2D\n",
    "# from keras.layers import BatchNormalization, Dropout, GlobalMaxPooling2D\n",
    "\n",
    "# def simple_cnn():\n",
    "#     pic_input = Input(shape=(75, 75, 3))\n",
    "#     ang_input = Input(shape=(1,))\n",
    "\n",
    "#     cnn = BatchNormalization()(pic_input)\n",
    "#     for i in range(4):\n",
    "#         cnn = Conv2D(8*2**i, kernel_size = (3,3), activation='relu')(cnn)\n",
    "#         cnn = MaxPooling2D((2,2))(cnn)\n",
    "#     cnn = GlobalMaxPooling2D()(cnn)\n",
    "#     cnn = concatenate([cnn,ang_input])\n",
    "#     cnn = Dense(32,activation='relu')(cnn)\n",
    "#     cnn = Dense(1, activation = 'sigmoid')(cnn)\n",
    "\n",
    "#     simple_cnn = Model(inputs=[pic_input,ang_input],outputs=cnn)\n",
    "\n",
    "#     simple_cnn.compile(optimizer='adam', loss = 'binary_crossentropy', metrics = ['accuracy'])\n",
    "#     return simple_cnn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Model\n",
    "from keras.layers import Input, Dense, Reshape, concatenate, Conv2D, Flatten, MaxPooling2D\n",
    "from keras.layers import BatchNormalization, Dropout, GlobalMaxPooling2D\n",
    "\n",
    "def getModel():\n",
    "    pic_input = Input(shape=(75, 75, 3))\n",
    "    ang_input = Input(shape=(1,))\n",
    "    \n",
    "    cnn = BatchNormalization()(pic_input)\n",
    "    cnn = Conv2D(64, kernel_size=(3, 3), activation='relu')(cnn)\n",
    "    cnn = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(cnn)\n",
    "    cnn = Dropout(0.3)(cnn)\n",
    "    cnn = Conv2D(128, kernel_size=(3, 3), activation='relu')(cnn)\n",
    "    cnn = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(cnn)\n",
    "    cnn = Dropout(0.3)(cnn)\n",
    "    cnn = Conv2D(128, kernel_size=(3, 3), activation='relu')(cnn)\n",
    "    cnn = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(cnn)\n",
    "    cnn = Dropout(0.3)(cnn)\n",
    "    cnn = Conv2D(64, kernel_size=(3, 3), activation='relu')(cnn)\n",
    "    cnn = MaxPooling2D(pool_size=(3, 3), strides=(2, 2))(cnn)\n",
    "    cnn = Dropout(0.3)(cnn)\n",
    "#     for i in range(4):\n",
    "#         cnn = Conv2D(8*2**i, kernel_size = (3,3), activation='relu')(cnn)\n",
    "#         cnn = MaxPooling2D((2,2))(cnn)\n",
    "#         cnn = Dropout(0.3)(cnn)\n",
    "    cnn = GlobalMaxPooling2D()(cnn)\n",
    "    cnn = concatenate([cnn,ang_input])\n",
    "    cnn = Dense(512,activation='relu')(cnn)\n",
    "    cnn = Dropout(0.3)(cnn)\n",
    "    cnn = Dense(256,activation='relu')(cnn)\n",
    "    cnn = Dropout(0.3)(cnn)\n",
    "    \n",
    "    predictions = Dense(1, activation = 'sigmoid')(cnn)\n",
    "\n",
    "    getModel = Model(inputs=[pic_input,ang_input],outputs=predictions)\n",
    "\n",
    "    adam = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=1e-6)\n",
    "    getModel.compile(optimizer= adam , loss = 'binary_crossentropy', metrics = ['accuracy'])\n",
    "    return getModel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "batch_size=64\n",
    "# Define the image transformations here\n",
    "gen = ImageDataGenerator(horizontal_flip = True,\n",
    "                         vertical_flip = True,\n",
    "                         width_shift_range = 0.,\n",
    "                         height_shift_range = 0.,\n",
    "                         channel_shift_range=0,\n",
    "                         zoom_range = 0.2,\n",
    "                         rotation_range = 10)\n",
    "\n",
    "# Here is the function that merges our two generators\n",
    "# We use the exact same generator with the same random seed for both the y and angle arrays\n",
    "def gen_flow_for_two_inputs(X1, X2, y):\n",
    "    genX1 = gen.flow(X1,y,  batch_size=batch_size,seed=666)\n",
    "    genX2 = gen.flow(X1,X2, batch_size=batch_size,seed=666)\n",
    "    while True:\n",
    "            X1i = genX1.next()\n",
    "            X2i = genX2.next()\n",
    "            #Assert arrays are equal - this was for peace of mind, but slows down training\n",
    "            #np.testing.assert_array_equal(X1i[0],X2i[0])\n",
    "            yield [X1i[0], X2i[1]], X1i[1]\n",
    "\n",
    "# Finally create generator\n",
    "gen_flow = gen_flow_for_two_inputs(X_train, X_angle_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_callbacks(filepath, patience=2):\n",
    "    es = EarlyStopping('val_loss', patience=patience, mode=\"min\")\n",
    "    msave = ModelCheckpoint(filepath, save_best_only=True)\n",
    "    return [es, msave]\n",
    "file_path = \".model_weights.hdf5\"\n",
    "callbacks = get_callbacks(filepath=file_path, patience=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Create the model\n",
    "model = simple_cnn()\n",
    "\n",
    "# Fit the model using our generator defined above\n",
    "histr=model.fit_generator(gen_flow, validation_data=([X_valid, X_angle_valid], y_valid),\n",
    "                    steps_per_epoch=len(X_train) / batch_size, epochs=0, callbacks=callbacks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_5 (InputLayer)            (None, 75, 75, 3)    0                                            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 75, 75, 3)    12          input_5[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 73, 73, 64)   1792        batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_9 (MaxPooling2D)  (None, 36, 36, 64)   0           conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dropout_13 (Dropout)            (None, 36, 36, 64)   0           max_pooling2d_9[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_10 (Conv2D)              (None, 34, 34, 128)  73856       dropout_13[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_10 (MaxPooling2D) (None, 16, 16, 128)  0           conv2d_10[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_14 (Dropout)            (None, 16, 16, 128)  0           max_pooling2d_10[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_11 (Conv2D)              (None, 14, 14, 128)  147584      dropout_14[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_11 (MaxPooling2D) (None, 6, 6, 128)    0           conv2d_11[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_15 (Dropout)            (None, 6, 6, 128)    0           max_pooling2d_11[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_12 (Conv2D)              (None, 4, 4, 64)     73792       dropout_15[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_12 (MaxPooling2D) (None, 1, 1, 64)     0           conv2d_12[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_16 (Dropout)            (None, 1, 1, 64)     0           max_pooling2d_12[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "global_max_pooling2d_3 (GlobalM (None, 64)           0           dropout_16[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "input_6 (InputLayer)            (None, 1)            0                                            \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_3 (Concatenate)     (None, 65)           0           global_max_pooling2d_3[0][0]     \n",
      "                                                                 input_6[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_7 (Dense)                 (None, 512)          33792       concatenate_3[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "dropout_17 (Dropout)            (None, 512)          0           dense_7[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_8 (Dense)                 (None, 256)          131328      dropout_17[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dropout_18 (Dropout)            (None, 256)          0           dense_8[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_9 (Dense)                 (None, 1)            257         dropout_18[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 462,413\n",
      "Trainable params: 462,407\n",
      "Non-trainable params: 6\n",
      "__________________________________________________________________________________________________\n",
      "Epoch 1/100\n",
      "24/24 [==============================] - 4s 168ms/step - loss: 1.0092 - acc: 0.5735 - val_loss: 0.8215 - val_acc: 0.5511\n",
      "Epoch 2/100\n",
      "24/24 [==============================] - 3s 133ms/step - loss: 0.6660 - acc: 0.6350 - val_loss: 0.5711 - val_acc: 0.7207\n",
      "Epoch 3/100\n",
      "24/24 [==============================] - 3s 135ms/step - loss: 0.5873 - acc: 0.6762 - val_loss: 0.6605 - val_acc: 0.6259\n",
      "Epoch 4/100\n",
      "24/24 [==============================] - 3s 140ms/step - loss: 0.5608 - acc: 0.6952 - val_loss: 0.6756 - val_acc: 0.5636\n",
      "Epoch 5/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.4638 - acc: 0.7635 - val_loss: 0.5038 - val_acc: 0.7307\n",
      "Epoch 6/100\n",
      "24/24 [==============================] - 3s 137ms/step - loss: 0.4669 - acc: 0.7858 - val_loss: 0.5124 - val_acc: 0.7332\n",
      "Epoch 7/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.4322 - acc: 0.7949 - val_loss: 0.5477 - val_acc: 0.6933\n",
      "Epoch 8/100\n",
      "24/24 [==============================] - 3s 139ms/step - loss: 0.4199 - acc: 0.8133 - val_loss: 0.4135 - val_acc: 0.7905\n",
      "Epoch 9/100\n",
      "24/24 [==============================] - 3s 139ms/step - loss: 0.4433 - acc: 0.8039 - val_loss: 0.4337 - val_acc: 0.8105\n",
      "Epoch 10/100\n",
      "24/24 [==============================] - 3s 134ms/step - loss: 0.4241 - acc: 0.8168 - val_loss: 0.5267 - val_acc: 0.7132\n",
      "Epoch 11/100\n",
      "24/24 [==============================] - 3s 140ms/step - loss: 0.3992 - acc: 0.8230 - val_loss: 0.4257 - val_acc: 0.7805\n",
      "Epoch 12/100\n",
      "24/24 [==============================] - 3s 133ms/step - loss: 0.4108 - acc: 0.8334 - val_loss: 0.4140 - val_acc: 0.8080\n",
      "Epoch 13/100\n",
      "24/24 [==============================] - 3s 140ms/step - loss: 0.3795 - acc: 0.8366 - val_loss: 0.3370 - val_acc: 0.8554\n",
      "Epoch 14/100\n",
      "24/24 [==============================] - 3s 135ms/step - loss: 0.3790 - acc: 0.8422 - val_loss: 0.3660 - val_acc: 0.8354\n",
      "Epoch 15/100\n",
      "24/24 [==============================] - 3s 135ms/step - loss: 0.3644 - acc: 0.8444 - val_loss: 0.4543 - val_acc: 0.7731\n",
      "Epoch 16/100\n",
      "24/24 [==============================] - 3s 139ms/step - loss: 0.3735 - acc: 0.8499 - val_loss: 0.5010 - val_acc: 0.7481\n",
      "Epoch 17/100\n",
      "24/24 [==============================] - 3s 141ms/step - loss: 0.4121 - acc: 0.8222 - val_loss: 0.4365 - val_acc: 0.7955\n",
      "Epoch 18/100\n",
      "24/24 [==============================] - 3s 134ms/step - loss: 0.4113 - acc: 0.8279 - val_loss: 0.3847 - val_acc: 0.8204\n",
      "Epoch 19/100\n",
      "24/24 [==============================] - 3s 137ms/step - loss: 0.3619 - acc: 0.8442 - val_loss: 0.4191 - val_acc: 0.7980\n",
      "Epoch 20/100\n",
      "24/24 [==============================] - 3s 139ms/step - loss: 0.3612 - acc: 0.8401 - val_loss: 0.4016 - val_acc: 0.8030\n",
      "Epoch 21/100\n",
      "24/24 [==============================] - 3s 135ms/step - loss: 0.3618 - acc: 0.8499 - val_loss: 0.3694 - val_acc: 0.8379\n",
      "Epoch 22/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.3438 - acc: 0.8574 - val_loss: 0.3283 - val_acc: 0.8529\n",
      "Epoch 23/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.3770 - acc: 0.8444 - val_loss: 0.3557 - val_acc: 0.8554\n",
      "Epoch 24/100\n",
      "24/24 [==============================] - 3s 137ms/step - loss: 0.3637 - acc: 0.8385 - val_loss: 0.3073 - val_acc: 0.8628\n",
      "Epoch 25/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.3545 - acc: 0.8496 - val_loss: 0.3134 - val_acc: 0.8529\n",
      "Epoch 26/100\n",
      "24/24 [==============================] - 3s 140ms/step - loss: 0.3405 - acc: 0.8569 - val_loss: 0.2950 - val_acc: 0.8678\n",
      "Epoch 27/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.3370 - acc: 0.8536 - val_loss: 0.3308 - val_acc: 0.8454\n",
      "Epoch 28/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.3303 - acc: 0.8488 - val_loss: 0.3175 - val_acc: 0.8579\n",
      "Epoch 29/100\n",
      "24/24 [==============================] - 3s 141ms/step - loss: 0.3533 - acc: 0.8589 - val_loss: 0.3104 - val_acc: 0.8628\n",
      "Epoch 30/100\n",
      "24/24 [==============================] - 3s 141ms/step - loss: 0.3484 - acc: 0.8484 - val_loss: 0.2622 - val_acc: 0.8878\n",
      "Epoch 31/100\n",
      "24/24 [==============================] - 3s 134ms/step - loss: 0.3084 - acc: 0.8696 - val_loss: 0.2655 - val_acc: 0.8878\n",
      "Epoch 32/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.3181 - acc: 0.8735 - val_loss: 0.2677 - val_acc: 0.8878\n",
      "Epoch 33/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.2982 - acc: 0.8794 - val_loss: 0.3027 - val_acc: 0.8554\n",
      "Epoch 34/100\n",
      "24/24 [==============================] - 3s 137ms/step - loss: 0.3214 - acc: 0.8623 - val_loss: 0.2572 - val_acc: 0.8803\n",
      "Epoch 35/100\n",
      "24/24 [==============================] - 3s 135ms/step - loss: 0.2824 - acc: 0.8862 - val_loss: 0.2224 - val_acc: 0.9052\n",
      "Epoch 36/100\n",
      "24/24 [==============================] - 3s 137ms/step - loss: 0.3278 - acc: 0.8563 - val_loss: 0.3162 - val_acc: 0.8479\n",
      "Epoch 37/100\n",
      "24/24 [==============================] - 3s 133ms/step - loss: 0.3120 - acc: 0.8646 - val_loss: 0.2582 - val_acc: 0.8778\n",
      "Epoch 38/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.3162 - acc: 0.8611 - val_loss: 0.2750 - val_acc: 0.8753\n",
      "Epoch 39/100\n",
      "24/24 [==============================] - 3s 135ms/step - loss: 0.2807 - acc: 0.8743 - val_loss: 0.2452 - val_acc: 0.8753\n",
      "Epoch 40/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.2829 - acc: 0.8836 - val_loss: 0.2696 - val_acc: 0.8753\n",
      "Epoch 41/100\n",
      "24/24 [==============================] - 3s 139ms/step - loss: 0.2981 - acc: 0.8719 - val_loss: 0.2358 - val_acc: 0.8878\n",
      "Epoch 42/100\n",
      "24/24 [==============================] - 3s 138ms/step - loss: 0.2879 - acc: 0.8828 - val_loss: 0.2334 - val_acc: 0.8953\n",
      "Epoch 43/100\n",
      "24/24 [==============================] - 3s 141ms/step - loss: 0.2949 - acc: 0.8686 - val_loss: 0.2447 - val_acc: 0.9027\n",
      "Epoch 44/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.2626 - acc: 0.8805 - val_loss: 0.2699 - val_acc: 0.8703\n",
      "Epoch 45/100\n",
      "24/24 [==============================] - 3s 136ms/step - loss: 0.2734 - acc: 0.8839 - val_loss: 0.2358 - val_acc: 0.8928\n"
     ]
    }
   ],
   "source": [
    "# Create the model\n",
    "model = getModel()\n",
    "model.summary()\n",
    "# Fit the model using our generator defined above\n",
    "histr=model.fit_generator(gen_flow, validation_data=([X_valid, X_angle_valid], y_valid),\n",
    "                    steps_per_epoch=24, epochs=100, callbacks=callbacks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss: 0.233633082679\n",
      "Train accuracy: 0.896093100681\n",
      "Test loss: 0.222421608364\n",
      "Test accuracy: 0.905236908177\n"
     ]
    }
   ],
   "source": [
    "model.load_weights(filepath=file_path)\n",
    "\n",
    "score = model.evaluate([X_train, X_angle_train], y_train, verbose=0)\n",
    "print('Train loss:', score[0])\n",
    "print('Train accuracy:', score[1])\n",
    "#Getting Test Score\n",
    "score = model.evaluate([X_valid, X_angle_valid], y_valid, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>is_iceberg</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6923</th>\n",
       "      <td>5c4ef9dd</td>\n",
       "      <td>0.594054</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4987</th>\n",
       "      <td>0dbd653f</td>\n",
       "      <td>0.136483</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7024</th>\n",
       "      <td>248be600</td>\n",
       "      <td>0.226744</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            id  is_iceberg\n",
       "6923  5c4ef9dd    0.594054\n",
       "4987  0dbd653f    0.136483\n",
       "7024  248be600    0.226744"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Predict on test data\n",
    "test_predictions = model.predict([X_test,X_angle_test])\n",
    "\n",
    "# Create .csv\n",
    "pred_df = test[['id']].copy()\n",
    "pred_df['is_iceberg'] = test_predictions\n",
    "pred_df.to_csv('model_01_data_DAG.csv', index = False)\n",
    "pred_df.sample(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Kaggle Score: 0.2454"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['loss', 'val_loss', 'val_acc', 'acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFTlYmJlZRQ4WZmZW0f8D/1qG+z2zstUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb830605eb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb8305cf828>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# list all data in history\n",
    "print(histr.history.keys())\n",
    "# summarize history for accuracy\n",
    "plt.plot(histr.history['acc'])\n",
    "plt.plot(histr.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "# summarize history for loss\n",
    "plt.plot(histr.history['loss'])\n",
    "plt.plot(histr.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}