Recognizing-CIFAR-10-images-Improved-Model-Data-Augmentation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Recognizing CIFAR-10 images with Data Augmentation (Part III)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "CIFAR-10 dataset contains 60,000 color images of size 32*32 pixels which are divided in 10 classes.\n",
    "We will use Keras library to import the dataset which contains 50,000 images as training iamges\n",
    "and 10,000 as test images\n",
    "\n",
    "In [Part I](https://github.com/chhayac/Machine-Learning-Notebooks/blob/master/Recognizing-CIFAR-10-images-Simple-Model.ipynb), \n",
    "we used a simple Convolutional Neural Network model with only one hidden layer. Model gave accuracy of 66.94% on test data. In [Part 2](https://github.com/chhayac/Machine-Learning-Notebooks/blob/master/Recognizing-CIFAR-10-images-Improved-Model.ipynb),we improved the accuracy further to 78.65 by increasing the number of convolutional layers.\n",
    "\n",
    "Goal of this part is to improve the accuracy of the model by using data augmentation like horizontal flip, zooming, etc."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from keras.datasets import cifar10\n",
    "from keras.utils import np_utils\n",
    "from keras.optimizers import SGD, Adam, RMSprop\n",
    "from keras.models import Sequential\n",
    "from keras.layers.convolutional import Conv2D, MaxPooling2D\n",
    "from keras.layers.core import Dense, Flatten, Activation, Dropout\n",
    "import matplotlib.pyplot as plt\n",
    "% matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "#load dataset\n",
    "(X_train, y_train), (X_test, y_test) = cifar10.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "BATCH_SIZE = 128\n",
    "nb_epochs = 40\n",
    "VALIDATION_SPLIT = 0.2\n",
    "num_classes = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_train, img_channels, img_rows, img_cols = X_train.shape\n",
    "num_test, _, _, _= X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train Shape: (50000, 32, 32, 3)\n",
      "Training samples: 50000\n",
      "Test Samples: 10000\n"
     ]
    }
   ],
   "source": [
    "print('X_train Shape:', X_train.shape)\n",
    "training_samples = X_train.shape[0]\n",
    "print(\"Training samples:\", training_samples)\n",
    "print(\"Test Samples:\", y_test.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_names =['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa9006e51d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize =(8, 3))\n",
    "for i in range(num_classes):\n",
    "    axis = fig.add_subplot(2, 5, 1+i, xticks = [], yticks = [])\n",
    "    idx = np.where(y_train[:] == i)[0]\n",
    "    features_idx = X_train[idx, ::]\n",
    "    image_num = np.random.randint(features_idx.shape[0])\n",
    "    image = features_idx[image_num,::]\n",
    "    axis.set_title(class_names[i])\n",
    "    plt.imshow(image)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "One hot Encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert to categorical\n",
    "y_train = np_utils.to_categorical(y_train, num_classes)\n",
    "y_test = np_utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "#float and normalization\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "X_train /= 255\n",
    "X_test /= 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "if K.backend() == 'tensorflow':\n",
    "    K.set_image_dim_ordering('tf')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Building sequential model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Network\n",
    "def create_cnn_model():\n",
    "    model = Sequential()\n",
    "    model.add(Conv2D(32, (3, 3), padding = 'same', input_shape=X_train.shape[1:]))\n",
    "    model.add(Activation('relu'))\n",
    "\n",
    "    # Adding more layers to improve the model\n",
    "    model.add(Conv2D(32, (3, 3), padding = 'same'))\n",
    "    model.add(Activation('relu'))\n",
    "    model.add(MaxPooling2D(pool_size= (2,2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    model.add(Conv2D(64, (3, 3), padding = 'same'))\n",
    "    model.add(Activation('relu'))\n",
    "    model.add(Conv2D(64, (3, 3)))\n",
    "    model.add(Activation('relu'))\n",
    "    model.add(MaxPooling2D(pool_size= (2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(512))\n",
    "    model.add(Activation('relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes))\n",
    "    model.add(Activation('softmax'))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_21 (Conv2D)           (None, 32, 32, 32)        896       \n",
      "_________________________________________________________________\n",
      "activation_31 (Activation)   (None, 32, 32, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_22 (Conv2D)           (None, 32, 32, 32)        9248      \n",
      "_________________________________________________________________\n",
      "activation_32 (Activation)   (None, 32, 32, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_11 (MaxPooling (None, 16, 16, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout_16 (Dropout)         (None, 16, 16, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_23 (Conv2D)           (None, 16, 16, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_33 (Activation)   (None, 16, 16, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_24 (Conv2D)           (None, 14, 14, 64)        36928     \n",
      "_________________________________________________________________\n",
      "activation_34 (Activation)   (None, 14, 14, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_12 (MaxPooling (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "dropout_17 (Dropout)         (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_6 (Flatten)          (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 512)               1606144   \n",
      "_________________________________________________________________\n",
      "activation_35 (Activation)   (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout_18 (Dropout)         (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 10)                5130      \n",
      "_________________________________________________________________\n",
      "activation_36 (Activation)   (None, 10)                0         \n",
      "=================================================================\n",
      "Total params: 1,676,842\n",
      "Trainable params: 1,676,842\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = create_cnn_model()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Compiling the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss= 'categorical_crossentropy', optimizer = 'rmsprop', metrics= ['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 40000 samples, validate on 10000 samples\n",
      "Epoch 1/40\n",
      "40000/40000 [==============================] - 13s 315us/step - loss: 1.7984 - acc: 0.3482 - val_loss: 1.4486 - val_acc: 0.4792\n",
      "Epoch 2/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 1.3739 - acc: 0.5104 - val_loss: 1.1942 - val_acc: 0.5778\n",
      "Epoch 3/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 1.1617 - acc: 0.5924 - val_loss: 1.0264 - val_acc: 0.6332\n",
      "Epoch 4/40\n",
      "40000/40000 [==============================] - 12s 305us/step - loss: 1.0202 - acc: 0.6408 - val_loss: 1.0277 - val_acc: 0.6481\n",
      "Epoch 5/40\n",
      "40000/40000 [==============================] - 12s 305us/step - loss: 0.9151 - acc: 0.6812 - val_loss: 0.8667 - val_acc: 0.6938\n",
      "Epoch 6/40\n",
      "40000/40000 [==============================] - 12s 305us/step - loss: 0.8339 - acc: 0.7088 - val_loss: 0.8095 - val_acc: 0.7215\n",
      "Epoch 7/40\n",
      "40000/40000 [==============================] - 12s 305us/step - loss: 0.7743 - acc: 0.7328 - val_loss: 0.7150 - val_acc: 0.7531\n",
      "Epoch 8/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.7214 - acc: 0.7482 - val_loss: 0.7398 - val_acc: 0.7457\n",
      "Epoch 9/40\n",
      "40000/40000 [==============================] - 12s 303us/step - loss: 0.6854 - acc: 0.7639 - val_loss: 0.7931 - val_acc: 0.7277\n",
      "Epoch 10/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.6486 - acc: 0.7768 - val_loss: 0.7630 - val_acc: 0.7533\n",
      "Epoch 11/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.6139 - acc: 0.7873 - val_loss: 0.7172 - val_acc: 0.7687\n",
      "Epoch 12/40\n",
      "40000/40000 [==============================] - 12s 302us/step - loss: 0.5945 - acc: 0.7965 - val_loss: 0.7446 - val_acc: 0.7536\n",
      "Epoch 13/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5773 - acc: 0.8019 - val_loss: 0.7163 - val_acc: 0.7699\n",
      "Epoch 14/40\n",
      "40000/40000 [==============================] - 12s 305us/step - loss: 0.5676 - acc: 0.8058 - val_loss: 0.7669 - val_acc: 0.7508\n",
      "Epoch 15/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5579 - acc: 0.8112 - val_loss: 0.7686 - val_acc: 0.7512\n",
      "Epoch 16/40\n",
      "40000/40000 [==============================] - 12s 303us/step - loss: 0.5442 - acc: 0.8149 - val_loss: 0.7752 - val_acc: 0.7633\n",
      "Epoch 17/40\n",
      "40000/40000 [==============================] - 12s 302us/step - loss: 0.5428 - acc: 0.8171 - val_loss: 0.8106 - val_acc: 0.7540\n",
      "Epoch 18/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5443 - acc: 0.8176 - val_loss: 0.6893 - val_acc: 0.7800\n",
      "Epoch 19/40\n",
      "40000/40000 [==============================] - 12s 301us/step - loss: 0.5364 - acc: 0.8238 - val_loss: 0.7981 - val_acc: 0.7587\n",
      "Epoch 20/40\n",
      "40000/40000 [==============================] - 12s 302us/step - loss: 0.5416 - acc: 0.8189 - val_loss: 0.8357 - val_acc: 0.7625\n",
      "Epoch 21/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5289 - acc: 0.8247 - val_loss: 0.7441 - val_acc: 0.7695\n",
      "Epoch 22/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5332 - acc: 0.8223 - val_loss: 0.8452 - val_acc: 0.7519\n",
      "Epoch 23/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5393 - acc: 0.8235 - val_loss: 0.6680 - val_acc: 0.7795\n",
      "Epoch 24/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5305 - acc: 0.8251 - val_loss: 0.6729 - val_acc: 0.7876\n",
      "Epoch 25/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5307 - acc: 0.8248 - val_loss: 0.7482 - val_acc: 0.7652\n",
      "Epoch 26/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5381 - acc: 0.8256 - val_loss: 0.9292 - val_acc: 0.7665\n",
      "Epoch 27/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5303 - acc: 0.8289 - val_loss: 0.6261 - val_acc: 0.7943\n",
      "Epoch 28/40\n",
      "40000/40000 [==============================] - 12s 305us/step - loss: 0.5447 - acc: 0.8225 - val_loss: 0.8138 - val_acc: 0.7681\n",
      "Epoch 29/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5460 - acc: 0.8243 - val_loss: 0.6878 - val_acc: 0.7805\n",
      "Epoch 30/40\n",
      "40000/40000 [==============================] - 12s 303us/step - loss: 0.5371 - acc: 0.8275 - val_loss: 0.6597 - val_acc: 0.7840\n",
      "Epoch 31/40\n",
      "40000/40000 [==============================] - 12s 301us/step - loss: 0.5312 - acc: 0.8269 - val_loss: 0.7262 - val_acc: 0.7862\n",
      "Epoch 32/40\n",
      "40000/40000 [==============================] - 12s 302us/step - loss: 0.5445 - acc: 0.8253 - val_loss: 0.6854 - val_acc: 0.7808\n",
      "Epoch 33/40\n",
      "40000/40000 [==============================] - 12s 302us/step - loss: 0.5407 - acc: 0.8250 - val_loss: 0.7579 - val_acc: 0.7715\n",
      "Epoch 34/40\n",
      "40000/40000 [==============================] - 12s 302us/step - loss: 0.5351 - acc: 0.8275 - val_loss: 0.6948 - val_acc: 0.7722\n",
      "Epoch 35/40\n",
      "40000/40000 [==============================] - 12s 301us/step - loss: 0.5449 - acc: 0.8264 - val_loss: 0.8917 - val_acc: 0.7873\n",
      "Epoch 36/40\n",
      "40000/40000 [==============================] - 12s 303us/step - loss: 0.5574 - acc: 0.8245 - val_loss: 0.6639 - val_acc: 0.7861\n",
      "Epoch 37/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5525 - acc: 0.8243 - val_loss: 0.9350 - val_acc: 0.7783\n",
      "Epoch 38/40\n",
      "40000/40000 [==============================] - 12s 303us/step - loss: 0.5548 - acc: 0.8231 - val_loss: 0.7403 - val_acc: 0.7728\n",
      "Epoch 39/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5566 - acc: 0.8219 - val_loss: 0.7684 - val_acc: 0.7856\n",
      "Epoch 40/40\n",
      "40000/40000 [==============================] - 12s 304us/step - loss: 0.5560 - acc: 0.8232 - val_loss: 0.8750 - val_acc: 0.7895\n"
     ]
    }
   ],
   "source": [
    "model_history = model.fit(X_train, y_train, batch_size = BATCH_SIZE, epochs = nb_epochs, \n",
    "          validation_split = VALIDATION_SPLIT, verbose = 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "List all data in history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['loss', 'val_loss', 'acc', 'val_acc'])\n"
     ]
    }
   ],
   "source": [
    "print(model_history.history.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Summarize history for accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa900274b38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_history.history['acc'])\n",
    "plt.plot(model_history.history['val_acc'])\n",
    "plt.title(\"Model Accuracy\")\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(['train','validation'], loc = 'upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Summarize history for loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa900253e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_history.history['loss'])\n",
    "plt.plot(model_history.history['val_loss'])\n",
    "plt.title(\"Model loss\")\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train','validation'], loc = 'upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 1s 97us/step\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(X_test, y_test, batch_size = BATCH_SIZE, verbose =1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Loss 0.915297679043\n",
      "Test accuracy 0.7836\n"
     ]
    }
   ],
   "source": [
    "print(\"Test Loss\", score[0])\n",
    "print(\"Test accuracy\", score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This accuracy is 12% better than what we got in Part I."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performing Data Augmentation\n",
    "Now we will perform data augmentation on the improved model which might improve the accuracy little more."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.image import ImageDataGenerator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load dataset\n",
    "#(X_train, y_train), (X_test, y_test) = cifar10.load_data()\n",
    "\n",
    "# Augmenting training set images\n",
    "datagen = ImageDataGenerator(zoom_range = 0.2,\n",
    "                            horizontal_flip = True)\n",
    "datagen.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_aug = create_cnn_model()\n",
    "model_aug.compile(loss = 'categorical_crossentropy', optimizer = 'rmsprop', metrics= ['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.callbacks import LearningRateScheduler, ModelCheckpoint, EarlyStopping\n",
    "import h5py as h5py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = 0.01\n",
    "def learning_rate_decay(epoch):\n",
    "    return lr * (0.1 ** int(epoch / 10)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/40\n",
      "391/390 [==============================] - 19s 48ms/step - loss: 1.7620 - acc: 0.3625 - val_loss: 1.3931 - val_acc: 0.5072\n",
      "Epoch 2/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 1.3146 - acc: 0.5350 - val_loss: 1.0961 - val_acc: 0.6049\n",
      "Epoch 3/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 1.1220 - acc: 0.6068 - val_loss: 0.9468 - val_acc: 0.6632\n",
      "Epoch 4/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 1.0077 - acc: 0.6486 - val_loss: 0.8146 - val_acc: 0.7159\n",
      "Epoch 5/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.9303 - acc: 0.6757 - val_loss: 0.8654 - val_acc: 0.6997\n",
      "Epoch 6/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.8822 - acc: 0.6940 - val_loss: 0.7934 - val_acc: 0.7249\n",
      "Epoch 7/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.8422 - acc: 0.7076 - val_loss: 0.8192 - val_acc: 0.7186\n",
      "Epoch 8/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.8130 - acc: 0.7196 - val_loss: 0.7421 - val_acc: 0.7434\n",
      "Epoch 9/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7976 - acc: 0.7256 - val_loss: 0.7139 - val_acc: 0.7624\n",
      "Epoch 10/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7759 - acc: 0.7321 - val_loss: 0.6575 - val_acc: 0.7818\n",
      "Epoch 11/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7548 - acc: 0.7417 - val_loss: 0.6609 - val_acc: 0.7769\n",
      "Epoch 12/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7528 - acc: 0.7419 - val_loss: 0.6139 - val_acc: 0.7959\n",
      "Epoch 13/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7390 - acc: 0.7487 - val_loss: 0.6693 - val_acc: 0.7727\n",
      "Epoch 14/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7391 - acc: 0.7482 - val_loss: 0.6527 - val_acc: 0.7827\n",
      "Epoch 15/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7350 - acc: 0.7508 - val_loss: 0.6920 - val_acc: 0.7713\n",
      "Epoch 16/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7355 - acc: 0.7526 - val_loss: 0.6900 - val_acc: 0.7683\n",
      "Epoch 17/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7294 - acc: 0.7527 - val_loss: 0.5973 - val_acc: 0.8040\n",
      "Epoch 18/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7343 - acc: 0.7516 - val_loss: 0.5944 - val_acc: 0.8033\n",
      "Epoch 19/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7270 - acc: 0.7578 - val_loss: 0.6038 - val_acc: 0.8083\n",
      "Epoch 20/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7320 - acc: 0.7560 - val_loss: 0.6144 - val_acc: 0.7983\n",
      "Epoch 21/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7240 - acc: 0.7588 - val_loss: 0.5837 - val_acc: 0.8075\n",
      "Epoch 22/40\n",
      "391/390 [==============================] - 18s 45ms/step - loss: 0.7374 - acc: 0.7537 - val_loss: 0.6393 - val_acc: 0.7938\n",
      "Epoch 23/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7400 - acc: 0.7549 - val_loss: 0.5803 - val_acc: 0.8079\n",
      "Epoch 24/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7341 - acc: 0.7573 - val_loss: 0.5777 - val_acc: 0.8157\n",
      "Epoch 25/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7374 - acc: 0.7545 - val_loss: 0.6797 - val_acc: 0.7887\n",
      "Epoch 26/40\n",
      "391/390 [==============================] - 18s 45ms/step - loss: 0.7382 - acc: 0.7563 - val_loss: 0.6107 - val_acc: 0.7999\n",
      "Epoch 27/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7393 - acc: 0.7556 - val_loss: 0.5978 - val_acc: 0.8078\n",
      "Epoch 28/40\n",
      "391/390 [==============================] - 18s 47ms/step - loss: 0.7417 - acc: 0.7542 - val_loss: 0.6044 - val_acc: 0.8101\n",
      "Epoch 29/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7394 - acc: 0.7540 - val_loss: 0.6143 - val_acc: 0.7980\n",
      "Epoch 30/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7412 - acc: 0.7558 - val_loss: 0.5862 - val_acc: 0.8102\n",
      "Epoch 31/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7446 - acc: 0.7530 - val_loss: 0.7044 - val_acc: 0.7918\n",
      "Epoch 32/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7540 - acc: 0.7519 - val_loss: 0.6096 - val_acc: 0.8074\n",
      "Epoch 33/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7472 - acc: 0.7540 - val_loss: 0.6090 - val_acc: 0.8126\n",
      "Epoch 34/40\n",
      "391/390 [==============================] - 18s 46ms/step - loss: 0.7572 - acc: 0.7511 - val_loss: 0.5873 - val_acc: 0.8073\n",
      "Epoch 00034: early stopping\n"
     ]
    }
   ],
   "source": [
    "model_history = model_aug.fit_generator(datagen.flow(X_train, y_train, \n",
    "                                                     batch_size = BATCH_SIZE), \n",
    "                                        steps_per_epoch = X_train.shape[0] / BATCH_SIZE,\n",
    "                                        epochs = nb_epochs, verbose = 1,\n",
    "                                        validation_data = (X_test, y_test),\n",
    "                                        callbacks = [ModelCheckpoint('model_aug.h5', save_best_only =True),\n",
    "                                                     EarlyStopping(monitor='val_acc', min_delta=0.000001, patience=10, \\\n",
    "                                                     verbose=1, mode='auto')])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Accuracy and Loss curves\n",
    "Summarize history for Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa903eadc18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_history.history['acc'])\n",
    "plt.plot(model_history.history['val_acc'])\n",
    "plt.title(\"Model Accuracy\")\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(['train','validation'], loc = 'upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Summarize history for Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa903deab70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model_history.history['loss'])\n",
    "plt.plot(model_history.history['val_loss'])\n",
    "plt.title(\"Model loss\")\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train','validation'], loc = 'upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 1s 99us/step\n"
     ]
    }
   ],
   "source": [
    "score = model_aug.evaluate(X_test, y_test, batch_size = BATCH_SIZE, verbose =1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Score 0.587267442322\n",
      "Test accuracy 0.8073\n"
     ]
    }
   ],
   "source": [
    "print(\"Test Loss\", score[0])\n",
    "print(\"Test accuracy\", score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clearly, by using data augmentation accuracy has improved by approximately 2%."
   ]
  }
 ],
 "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
}