unit10_cnn/demo4_classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zJDFJVzRHvIO"
   },
   "source": [
    "# Demo:  CNN Classifier for CIFAR10\n",
    "\n",
    "In this demo, you will learn to:\n",
    "\n",
    "* Load the classic CIFAR10 dataset from keras and visualize the images\n",
    "* Train and test a simple CNN classifier for the dataset\n",
    "* Enhance the classifiers with batch normalization, dropout and data augmentation and evaluate the relative performance gains.\n",
    "\n",
    "\n",
    "The [CIFAR-10 dataset](https://www.cs.toronto.edu/~kriz/cifar.html) is a collection of 60,000 color, `32x32`-pixel images in ten classes. Classes include common objects such as airplanes, automobiles, birds, cats and so on. There are 50,000 train and 10,000 test images. Keras can automatically download the dataset from `keras.datasets`. Note that it will take some time to downloading the dataset for the first time. But, the download will be saved in cache and would not be needed to download again. \n",
    "\n",
    "State of the art results are achieved using very large Convolutional Neural networks. Model performance is reported in this [classification accuracy table](http://rodrigob.github.io/are_we_there_yet/build/classification_datasets_results.html#43494641522d3130), with state-of-the-art results at 96.5% with human performance on the problem at 94%.  In this demo, we will use a very basic CNN.  With suitable enhancements we can performance ~78%.\n",
    "\n",
    "Most of the code and the description in this demo is taken from `cifar10_cnn.py` available at [keras-team Github page](https://github.com/keras-team/keras/blob/master/examples/cifar10_cnn.py) as well as from the tutorial page by [Jason Brownlee](https://machinelearningmastery.com/object-recognition-convolutional-neural-networks-keras-deep-learning-library/).  Also, thanks to [Phil Schniter](http://www2.ece.ohio-state.edu/~schniter/) for helping adjust some parameters.\n",
    "\n",
    "\n",
    "## Loading Basic Packages\n",
    "\n",
    "We first load some basic packages."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "y1v-qtceHvIP"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "matplotlib.rcParams.update({'font.size': 16})\n",
    "\n",
    "from tensorflow.keras.datasets import cifar10\n",
    "import pickle"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "XKBgYgqhHvIP"
   },
   "source": [
    "This demo requires use of a GPU.  Otherwise, the fitting will be far too slow.  The following function is one way to test if your machine is using a GPU."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "w1DAYR1_HvIQ",
    "outputId": "45fc106e-d6ea-49e9-8d76-b2512376cea4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[name: \"/device:CPU:0\"\n",
      "device_type: \"CPU\"\n",
      "memory_limit: 268435456\n",
      "locality {\n",
      "}\n",
      "incarnation: 387720724492694467\n",
      ", name: \"/device:GPU:0\"\n",
      "device_type: \"GPU\"\n",
      "memory_limit: 15703311680\n",
      "locality {\n",
      "  bus_id: 1\n",
      "  links {\n",
      "  }\n",
      "}\n",
      "incarnation: 8126079203260770633\n",
      "physical_device_desc: \"device: 0, name: Tesla P100-PCIE-16GB, pci bus id: 0000:00:04.0, compute capability: 6.0\"\n",
      "]\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.python.client import device_lib\n",
    "print(device_lib.list_local_devices())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "e8IPwvyyHvIQ"
   },
   "source": [
    "Create a 3×3 plot of photographs. The images have been scaled up from their small 32×32 size, but you can clearly see trucks horses and cars. You can also see some distortion in some images that have been forced to the square aspect ratio."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "hUjgT85jHvIR",
    "outputId": "86ebc00c-23d6-410e-b361-68c9602b7919"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Xtr shape:  (50000, 32, 32, 3)\n",
      "Xts shape:  (10000, 32, 32, 3)\n"
     ]
    }
   ],
   "source": [
    "# load data\n",
    "(Xtr,ytr), (Xts,yts) = cifar10.load_data()\n",
    "ntr, nrow, ncol, nchan = Xtr.shape\n",
    "nts = Xts.shape[0]\n",
    "\n",
    "print('Xtr shape:  ' + str(Xtr.shape))\n",
    "print('Xts shape:  ' + str(Xts.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 252
    },
    "id": "iMYYvAWdHvIR",
    "outputId": "0298b9b8-b883-4152-b921-58a936731def",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plt_image(im):\n",
    "    plt.imshow(im)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    \n",
    "# Print a few random samples\n",
    "nplot = 9\n",
    "I = np.random.permutation(ntr)\n",
    "for i in range(0, 9):\n",
    "    plt.subplot(3,3,i+1)\n",
    "    plt_image(Xtr[I[i]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "anGr88r1HvIR"
   },
   "source": [
    "import some more packages for building CNN model and saving the trained model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "id": "Hy0x6QfTHvIR"
   },
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.models import load_model #save and load models\n",
    "from tensorflow.keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D\n",
    "from tensorflow.keras.layers import BatchNormalization\n",
    "from tensorflow.keras import optimizers\n",
    "import tensorflow.keras.backend as K"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xC4bgrJqHvIR"
   },
   "source": [
    "The pixel values are in the range of 0 to 255 for each of the red, green and blue channels.\n",
    "\n",
    "It is good practice to work with normalized data. Because the input values are well understood, we can easily normalize to the range 0 to 1 by dividing each value by the maximum observation which is 255.  Note, the data is loaded as integers, so we must cast it to floating point values in order to perform the division."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "4pAbmnKoHvIS"
   },
   "outputs": [],
   "source": [
    "Xtr = Xtr.astype('float32') / 255.\n",
    "Xts = Xts.astype('float32') / 255."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "cD_6YIhpHvIS"
   },
   "source": [
    "## Testing Different Classifiers\n",
    "\n",
    "We now define a function to create a model.  The function has two paramters:\n",
    "\n",
    "* `use_bn`:  Adds BatchNormalization\n",
    "* `use_dropout`:  Adds Dropout layers\n",
    "\n",
    "By setting the parameters, we can experiment with different model features and compare their performance.  The model has two convolutional layers + two FC layers.  Dropout, if added, is done on the FC layers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "id": "z76GxgyJHvIS"
   },
   "outputs": [],
   "source": [
    "def create_mod(use_dropout=False, use_bn=False):\n",
    "    num_classes = 10\n",
    "    model = Sequential()\n",
    "    model.add(Conv2D(32, (3, 3), \n",
    "                     padding='valid', activation='relu',\n",
    "                     input_shape=Xtr.shape[1:]))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    if use_bn:\n",
    "        model.add(BatchNormalization())\n",
    "    model.add(Conv2D(32, (3, 3), activation='relu'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Flatten())\n",
    "    if use_bn:\n",
    "        model.add(BatchNormalization())\n",
    "    if use_dropout:\n",
    "        model.add(Dropout(0.5))\n",
    "    model.add(Dense(512, activation = 'relu'))\n",
    "    if use_bn:\n",
    "        model.add(BatchNormalization())\n",
    "    if use_dropout:\n",
    "        model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation='softmax'))\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "50YGcvj2HvIS"
   },
   "source": [
    "To study data augmentation, we also create an `ImageDataGenerator` object that will create augmented images for the training data set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "Qz4CPfK8HvIS"
   },
   "outputs": [],
   "source": [
    "def create_datagen():\n",
    "    datagen = ImageDataGenerator(\n",
    "        featurewise_center=False,  # set input mean to 0 over the dataset\n",
    "        samplewise_center=False,  # set each sample mean to 0\n",
    "        featurewise_std_normalization=False,  # divide inputs by std of the dataset\n",
    "        samplewise_std_normalization=False,  # divide each input by its std\n",
    "        zca_whitening=False,  # apply ZCA whitening\n",
    "        zca_epsilon=1e-06,  # epsilon for ZCA whitening\n",
    "        rotation_range=0,  # randomly rotate images in the range (degrees, 0 to 180)\n",
    "        # randomly shift images horizontally (fraction of total width)\n",
    "        width_shift_range=0.1,\n",
    "        # randomly shift images vertically (fraction of total height)\n",
    "        height_shift_range=0.1,\n",
    "        shear_range=0.,  # set range for random shear\n",
    "        zoom_range=0.,  # set range for random zoom\n",
    "        channel_shift_range=0.,  # set range for random channel shifts\n",
    "        # set mode for filling points outside the input boundaries\n",
    "        fill_mode='nearest',\n",
    "        cval=0.,  # value used for fill_mode = \"constant\"\n",
    "        horizontal_flip=True,  # randomly flip images\n",
    "        vertical_flip=False,  # randomly flip images\n",
    "        # set rescaling factor (applied before any other transformation)\n",
    "        rescale=None,\n",
    "        # set function that will be applied on each input\n",
    "        preprocessing_function=None,\n",
    "        # image data format, either \"channels_first\" or \"channels_last\"\n",
    "        data_format=None,\n",
    "        # fraction of images reserved for validation (strictly between 0 and 1)\n",
    "        validation_split=0.0)\n",
    "    return datagen\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IWZqT4ey5saR"
   },
   "source": [
    "Convert labels to categorical."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "gEEw2P1L5uX3"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.utils import to_categorical\n",
    "n_classes = 10\n",
    "ytr_cat = to_categorical(ytr, n_classes)\n",
    "yts_cat = to_categorical(yts, n_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "izii_SEcHvIS"
   },
   "source": [
    "Now, we run over all possible options:  The options are:\n",
    "        \n",
    "* `basic`:  Basic CNN, no batch normalization or dropout   \n",
    "* `bn`:  Basic CNN + batch normalization\n",
    "* `dropout`:  Basic CNN + batch normalization + dropout\n",
    "* `dataaug`:  Basic CNN + batch normalization + dropout + data augmentation\n",
    "\n",
    "This will take tens of minutes per model even on a GPU.  We save two files for each run: An `h5` model file and pickle file with the accuracy history."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "hIgB7AGiHvIS",
    "outputId": "29efbefc-4aeb-4d56-c4ba-e03569b82077"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 30, 30, 32)        896       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 13, 13, 32)        9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 512)               590336    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 605,610\n",
      "Trainable params: 605,610\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Epoch 1/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 1.6162 - accuracy: 0.4103 - val_loss: 1.1557 - val_accuracy: 0.5896\n",
      "Epoch 2/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.0668 - accuracy: 0.6231 - val_loss: 0.9813 - val_accuracy: 0.6513\n",
      "Epoch 3/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.8811 - accuracy: 0.6923 - val_loss: 0.9094 - val_accuracy: 0.6863\n",
      "Epoch 4/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.7448 - accuracy: 0.7364 - val_loss: 0.8716 - val_accuracy: 0.7001\n",
      "Epoch 5/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.6285 - accuracy: 0.7830 - val_loss: 0.8535 - val_accuracy: 0.7044\n",
      "Epoch 6/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.5253 - accuracy: 0.8212 - val_loss: 0.9186 - val_accuracy: 0.7004\n",
      "Epoch 7/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.4358 - accuracy: 0.8504 - val_loss: 0.9208 - val_accuracy: 0.7076\n",
      "Epoch 8/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.3487 - accuracy: 0.8832 - val_loss: 0.9675 - val_accuracy: 0.7076\n",
      "Epoch 9/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.2733 - accuracy: 0.9127 - val_loss: 1.0555 - val_accuracy: 0.7097\n",
      "Epoch 10/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.2071 - accuracy: 0.9359 - val_loss: 1.1383 - val_accuracy: 0.6977\n",
      "Epoch 11/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.1463 - accuracy: 0.9598 - val_loss: 1.1966 - val_accuracy: 0.7105\n",
      "Epoch 12/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.1089 - accuracy: 0.9712 - val_loss: 1.2996 - val_accuracy: 0.7065\n",
      "Epoch 13/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0786 - accuracy: 0.9820 - val_loss: 1.3889 - val_accuracy: 0.7072\n",
      "Epoch 14/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0555 - accuracy: 0.9887 - val_loss: 1.5173 - val_accuracy: 0.7074\n",
      "Epoch 15/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0383 - accuracy: 0.9932 - val_loss: 1.6267 - val_accuracy: 0.7045\n",
      "Epoch 16/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0277 - accuracy: 0.9960 - val_loss: 1.7118 - val_accuracy: 0.7046\n",
      "Epoch 17/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0216 - accuracy: 0.9966 - val_loss: 1.8392 - val_accuracy: 0.7043\n",
      "Epoch 18/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0169 - accuracy: 0.9977 - val_loss: 1.8673 - val_accuracy: 0.7053\n",
      "Epoch 19/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0127 - accuracy: 0.9987 - val_loss: 1.9765 - val_accuracy: 0.7002\n",
      "Epoch 20/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0107 - accuracy: 0.9984 - val_loss: 2.0373 - val_accuracy: 0.7036\n",
      "Epoch 21/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0066 - accuracy: 0.9995 - val_loss: 2.0810 - val_accuracy: 0.6985\n",
      "Epoch 22/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0097 - accuracy: 0.9985 - val_loss: 2.2152 - val_accuracy: 0.6992\n",
      "Epoch 23/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0151 - accuracy: 0.9966 - val_loss: 2.1736 - val_accuracy: 0.7060\n",
      "Epoch 24/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0033 - accuracy: 0.9994 - val_loss: 2.2197 - val_accuracy: 0.7011\n",
      "Epoch 25/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0057 - accuracy: 0.9993 - val_loss: 2.2662 - val_accuracy: 0.7059\n",
      "Epoch 26/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0018 - accuracy: 0.9999 - val_loss: 2.5503 - val_accuracy: 0.6819\n",
      "Epoch 27/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0172 - accuracy: 0.9948 - val_loss: 2.3050 - val_accuracy: 0.7080\n",
      "Epoch 28/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0014 - accuracy: 1.0000 - val_loss: 2.3746 - val_accuracy: 0.7085\n",
      "Epoch 29/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0137 - accuracy: 0.9963 - val_loss: 2.4112 - val_accuracy: 0.7028\n",
      "Epoch 30/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0014 - accuracy: 1.0000 - val_loss: 2.4311 - val_accuracy: 0.7085\n",
      "Epoch 31/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0035 - accuracy: 0.9991 - val_loss: 2.4197 - val_accuracy: 0.7061\n",
      "Epoch 32/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0014 - accuracy: 0.9999 - val_loss: 2.4601 - val_accuracy: 0.7084\n",
      "Epoch 33/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 5.0774e-04 - accuracy: 1.0000 - val_loss: 2.5247 - val_accuracy: 0.7104\n",
      "Epoch 34/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 3.4881e-04 - accuracy: 1.0000 - val_loss: 2.6164 - val_accuracy: 0.7108\n",
      "Epoch 35/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0131 - accuracy: 0.9967 - val_loss: 2.5448 - val_accuracy: 0.7091\n",
      "Epoch 36/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 5.8406e-04 - accuracy: 1.0000 - val_loss: 2.5838 - val_accuracy: 0.7089\n",
      "Epoch 37/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 3.2606e-04 - accuracy: 1.0000 - val_loss: 2.6388 - val_accuracy: 0.7096\n",
      "Epoch 38/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 2.3605e-04 - accuracy: 1.0000 - val_loss: 2.7011 - val_accuracy: 0.7088\n",
      "Epoch 39/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 9.8038e-04 - accuracy: 0.9998 - val_loss: 2.7728 - val_accuracy: 0.6999\n",
      "Epoch 40/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0023 - accuracy: 0.9997 - val_loss: 2.6758 - val_accuracy: 0.7094\n",
      "Epoch 41/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.7154e-04 - accuracy: 1.0000 - val_loss: 2.7150 - val_accuracy: 0.7092\n",
      "Epoch 42/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.8747e-04 - accuracy: 1.0000 - val_loss: 2.7626 - val_accuracy: 0.7090\n",
      "Epoch 43/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.3597e-04 - accuracy: 1.0000 - val_loss: 2.8312 - val_accuracy: 0.7096\n",
      "Epoch 44/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0075 - accuracy: 0.9977 - val_loss: 2.7925 - val_accuracy: 0.7056\n",
      "Epoch 45/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 3.0564e-04 - accuracy: 1.0000 - val_loss: 2.7922 - val_accuracy: 0.7085\n",
      "Epoch 46/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.5481e-04 - accuracy: 1.0000 - val_loss: 2.8276 - val_accuracy: 0.7095\n",
      "Epoch 47/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.1263e-04 - accuracy: 1.0000 - val_loss: 2.8802 - val_accuracy: 0.7100\n",
      "Epoch 48/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 8.6365e-05 - accuracy: 1.0000 - val_loss: 2.9452 - val_accuracy: 0.7101\n",
      "Epoch 49/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 0.0029 - accuracy: 0.9990 - val_loss: 2.9107 - val_accuracy: 0.7055\n",
      "Epoch 50/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 2.5717e-04 - accuracy: 1.0000 - val_loss: 2.9184 - val_accuracy: 0.7073\n",
      "Epoch 51/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.1772e-04 - accuracy: 1.0000 - val_loss: 2.9447 - val_accuracy: 0.7076\n",
      "Epoch 52/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 8.8131e-05 - accuracy: 1.0000 - val_loss: 2.9757 - val_accuracy: 0.7081\n",
      "Epoch 53/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 6.3620e-05 - accuracy: 1.0000 - val_loss: 3.0171 - val_accuracy: 0.7077\n",
      "Epoch 54/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 4.7761e-05 - accuracy: 1.0000 - val_loss: 3.0711 - val_accuracy: 0.7092\n",
      "Epoch 55/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0046 - accuracy: 0.9988 - val_loss: 3.0155 - val_accuracy: 0.7088\n",
      "Epoch 56/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.0728e-04 - accuracy: 1.0000 - val_loss: 3.0280 - val_accuracy: 0.7095\n",
      "Epoch 57/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 7.0703e-05 - accuracy: 1.0000 - val_loss: 3.0562 - val_accuracy: 0.7102\n",
      "Epoch 58/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 4.9259e-05 - accuracy: 1.0000 - val_loss: 3.0918 - val_accuracy: 0.7096\n",
      "Epoch 59/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 3.4065e-04 - accuracy: 0.9999 - val_loss: 3.1163 - val_accuracy: 0.6990\n",
      "Epoch 60/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0021 - accuracy: 0.9994 - val_loss: 3.0515 - val_accuracy: 0.7071\n",
      "Epoch 61/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.0347e-04 - accuracy: 1.0000 - val_loss: 3.0613 - val_accuracy: 0.7111\n",
      "Epoch 62/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 6.7735e-05 - accuracy: 1.0000 - val_loss: 3.0745 - val_accuracy: 0.7105\n",
      "Epoch 63/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 4.8937e-05 - accuracy: 1.0000 - val_loss: 3.1030 - val_accuracy: 0.7120\n",
      "Epoch 64/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 3.6386e-05 - accuracy: 1.0000 - val_loss: 3.1479 - val_accuracy: 0.7107\n",
      "Epoch 65/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.8121e-05 - accuracy: 1.0000 - val_loss: 3.2013 - val_accuracy: 0.7104\n",
      "Epoch 66/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.2160e-05 - accuracy: 1.0000 - val_loss: 3.2729 - val_accuracy: 0.7101\n",
      "Epoch 67/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 6.9303e-04 - accuracy: 0.9998 - val_loss: 3.2801 - val_accuracy: 0.7036\n",
      "Epoch 68/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.1627e-04 - accuracy: 1.0000 - val_loss: 3.2636 - val_accuracy: 0.7080\n",
      "Epoch 69/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 4.4622e-05 - accuracy: 1.0000 - val_loss: 3.2677 - val_accuracy: 0.7091\n",
      "Epoch 70/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 3.0537e-05 - accuracy: 1.0000 - val_loss: 3.2811 - val_accuracy: 0.7096\n",
      "Epoch 71/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.1528e-05 - accuracy: 1.0000 - val_loss: 3.3053 - val_accuracy: 0.7109\n",
      "Epoch 72/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.6068e-05 - accuracy: 1.0000 - val_loss: 3.3367 - val_accuracy: 0.7097\n",
      "Epoch 73/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.2657e-05 - accuracy: 1.0000 - val_loss: 3.3878 - val_accuracy: 0.7111\n",
      "Epoch 74/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0023 - accuracy: 0.9993 - val_loss: 3.3556 - val_accuracy: 0.7073\n",
      "Epoch 75/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 4.9636e-05 - accuracy: 1.0000 - val_loss: 3.3569 - val_accuracy: 0.7079\n",
      "Epoch 76/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.9458e-05 - accuracy: 1.0000 - val_loss: 3.3599 - val_accuracy: 0.7091\n",
      "Epoch 77/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.0514e-05 - accuracy: 1.0000 - val_loss: 3.3710 - val_accuracy: 0.7107\n",
      "Epoch 78/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.5008e-05 - accuracy: 1.0000 - val_loss: 3.3855 - val_accuracy: 0.7099\n",
      "Epoch 79/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.1849e-05 - accuracy: 1.0000 - val_loss: 3.4183 - val_accuracy: 0.7099\n",
      "Epoch 80/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 9.3238e-06 - accuracy: 1.0000 - val_loss: 3.4584 - val_accuracy: 0.7088\n",
      "Epoch 81/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 8.0255e-04 - accuracy: 0.9997 - val_loss: 3.4772 - val_accuracy: 0.7057\n",
      "Epoch 82/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 8.4669e-05 - accuracy: 1.0000 - val_loss: 3.4562 - val_accuracy: 0.7070\n",
      "Epoch 83/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.8780e-05 - accuracy: 1.0000 - val_loss: 3.4499 - val_accuracy: 0.7078\n",
      "Epoch 84/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.7757e-05 - accuracy: 1.0000 - val_loss: 3.4554 - val_accuracy: 0.7081\n",
      "Epoch 85/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.3394e-05 - accuracy: 1.0000 - val_loss: 3.4688 - val_accuracy: 0.7073\n",
      "Epoch 86/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 9.9492e-06 - accuracy: 1.0000 - val_loss: 3.4849 - val_accuracy: 0.7082\n",
      "Epoch 87/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 7.7466e-06 - accuracy: 1.0000 - val_loss: 3.5142 - val_accuracy: 0.7085\n",
      "Epoch 88/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 5.9448e-06 - accuracy: 1.0000 - val_loss: 3.5643 - val_accuracy: 0.7093\n",
      "Epoch 89/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 4.7104e-06 - accuracy: 1.0000 - val_loss: 3.6178 - val_accuracy: 0.7088\n",
      "Epoch 90/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 3.8524e-06 - accuracy: 1.0000 - val_loss: 3.6961 - val_accuracy: 0.7084\n",
      "Epoch 91/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 0.0011 - accuracy: 0.9997 - val_loss: 3.6964 - val_accuracy: 0.7069\n",
      "Epoch 92/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 2.1237e-05 - accuracy: 1.0000 - val_loss: 3.6961 - val_accuracy: 0.7071\n",
      "Epoch 93/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.1315e-05 - accuracy: 1.0000 - val_loss: 3.6990 - val_accuracy: 0.7063\n",
      "Epoch 94/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 7.7488e-06 - accuracy: 1.0000 - val_loss: 3.7042 - val_accuracy: 0.7070\n",
      "Epoch 95/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 5.6097e-06 - accuracy: 1.0000 - val_loss: 3.7158 - val_accuracy: 0.7079\n",
      "Epoch 96/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 4.1591e-06 - accuracy: 1.0000 - val_loss: 3.7281 - val_accuracy: 0.7092\n",
      "Epoch 97/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 3.0244e-06 - accuracy: 1.0000 - val_loss: 3.7514 - val_accuracy: 0.7077\n",
      "Epoch 98/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 2.4895e-06 - accuracy: 1.0000 - val_loss: 3.7923 - val_accuracy: 0.7079\n",
      "Epoch 99/100\n",
      "1563/1563 [==============================] - 7s 4ms/step - loss: 1.9829e-06 - accuracy: 1.0000 - val_loss: 3.8378 - val_accuracy: 0.7089\n",
      "Epoch 100/100\n",
      "1563/1563 [==============================] - 6s 4ms/step - loss: 1.6676e-06 - accuracy: 1.0000 - val_loss: 3.8937 - val_accuracy: 0.7085\n",
      "Model saved as cifar_basic.h5\n",
      "History saved as hist_basic.p\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 30, 30, 32)        896       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 15, 15, 32)        128       \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 13, 13, 32)        9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 1152)              4608      \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 512)               590336    \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 512)               2048      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 612,394\n",
      "Trainable params: 609,002\n",
      "Non-trainable params: 3,392\n",
      "_________________________________________________________________\n",
      "None\n",
      "Epoch 1/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 1.5799 - accuracy: 0.4694 - val_loss: 1.5007 - val_accuracy: 0.5162\n",
      "Epoch 2/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.9625 - accuracy: 0.6635 - val_loss: 0.9366 - val_accuracy: 0.6800\n",
      "Epoch 3/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.7662 - accuracy: 0.7338 - val_loss: 0.8939 - val_accuracy: 0.6922\n",
      "Epoch 4/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.6251 - accuracy: 0.7802 - val_loss: 0.9233 - val_accuracy: 0.6886\n",
      "Epoch 5/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.5006 - accuracy: 0.8292 - val_loss: 0.9518 - val_accuracy: 0.6917\n",
      "Epoch 6/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.3910 - accuracy: 0.8673 - val_loss: 0.9184 - val_accuracy: 0.7143\n",
      "Epoch 7/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.3141 - accuracy: 0.8965 - val_loss: 0.9725 - val_accuracy: 0.7092\n",
      "Epoch 8/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.2459 - accuracy: 0.9189 - val_loss: 0.9782 - val_accuracy: 0.7180\n",
      "Epoch 9/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.2027 - accuracy: 0.9341 - val_loss: 1.0549 - val_accuracy: 0.7104\n",
      "Epoch 10/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.1608 - accuracy: 0.9493 - val_loss: 1.0527 - val_accuracy: 0.7111\n",
      "Epoch 11/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.1385 - accuracy: 0.9559 - val_loss: 1.0718 - val_accuracy: 0.7181\n",
      "Epoch 12/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.1150 - accuracy: 0.9631 - val_loss: 1.0919 - val_accuracy: 0.7177\n",
      "Epoch 13/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.1028 - accuracy: 0.9675 - val_loss: 1.1651 - val_accuracy: 0.7193\n",
      "Epoch 14/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0932 - accuracy: 0.9703 - val_loss: 1.2165 - val_accuracy: 0.7170\n",
      "Epoch 15/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0770 - accuracy: 0.9760 - val_loss: 1.1993 - val_accuracy: 0.7179\n",
      "Epoch 16/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0749 - accuracy: 0.9762 - val_loss: 1.2156 - val_accuracy: 0.7223\n",
      "Epoch 17/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0665 - accuracy: 0.9788 - val_loss: 1.2677 - val_accuracy: 0.7160\n",
      "Epoch 18/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0575 - accuracy: 0.9822 - val_loss: 1.2951 - val_accuracy: 0.7171\n",
      "Epoch 19/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0536 - accuracy: 0.9835 - val_loss: 1.2442 - val_accuracy: 0.7254\n",
      "Epoch 20/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0490 - accuracy: 0.9857 - val_loss: 1.3404 - val_accuracy: 0.7190\n",
      "Epoch 21/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0496 - accuracy: 0.9838 - val_loss: 1.3129 - val_accuracy: 0.7226\n",
      "Epoch 22/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0429 - accuracy: 0.9871 - val_loss: 1.3309 - val_accuracy: 0.7229\n",
      "Epoch 23/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0426 - accuracy: 0.9873 - val_loss: 1.3873 - val_accuracy: 0.7203\n",
      "Epoch 24/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0427 - accuracy: 0.9869 - val_loss: 1.3193 - val_accuracy: 0.7293\n",
      "Epoch 25/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0385 - accuracy: 0.9881 - val_loss: 1.4077 - val_accuracy: 0.7219\n",
      "Epoch 26/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0319 - accuracy: 0.9903 - val_loss: 1.3591 - val_accuracy: 0.7270\n",
      "Epoch 27/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0315 - accuracy: 0.9906 - val_loss: 1.4571 - val_accuracy: 0.7160\n",
      "Epoch 28/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0294 - accuracy: 0.9911 - val_loss: 1.4130 - val_accuracy: 0.7241\n",
      "Epoch 29/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0311 - accuracy: 0.9899 - val_loss: 1.4340 - val_accuracy: 0.7205\n",
      "Epoch 30/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0281 - accuracy: 0.9914 - val_loss: 1.5320 - val_accuracy: 0.7132\n",
      "Epoch 31/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0231 - accuracy: 0.9929 - val_loss: 1.4868 - val_accuracy: 0.7195\n",
      "Epoch 32/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0280 - accuracy: 0.9911 - val_loss: 1.4396 - val_accuracy: 0.7296\n",
      "Epoch 33/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0266 - accuracy: 0.9920 - val_loss: 1.4881 - val_accuracy: 0.7183\n",
      "Epoch 34/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0215 - accuracy: 0.9938 - val_loss: 1.4431 - val_accuracy: 0.7299\n",
      "Epoch 35/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0202 - accuracy: 0.9945 - val_loss: 1.4562 - val_accuracy: 0.7267\n",
      "Epoch 36/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.0217 - accuracy: 0.9934 - val_loss: 1.4816 - val_accuracy: 0.7250\n",
      "Epoch 37/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0219 - accuracy: 0.9929 - val_loss: 1.4276 - val_accuracy: 0.7256\n",
      "Epoch 38/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0206 - accuracy: 0.9936 - val_loss: 1.4467 - val_accuracy: 0.7348\n",
      "Epoch 39/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0195 - accuracy: 0.9944 - val_loss: 1.4614 - val_accuracy: 0.7305\n",
      "Epoch 40/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0189 - accuracy: 0.9939 - val_loss: 1.4712 - val_accuracy: 0.7336\n",
      "Epoch 41/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0170 - accuracy: 0.9952 - val_loss: 1.4967 - val_accuracy: 0.7295\n",
      "Epoch 42/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0173 - accuracy: 0.9948 - val_loss: 1.4471 - val_accuracy: 0.7313\n",
      "Epoch 43/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0171 - accuracy: 0.9949 - val_loss: 1.5811 - val_accuracy: 0.7162\n",
      "Epoch 44/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0184 - accuracy: 0.9949 - val_loss: 1.5029 - val_accuracy: 0.7327\n",
      "Epoch 45/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0174 - accuracy: 0.9949 - val_loss: 1.6000 - val_accuracy: 0.7261\n",
      "Epoch 46/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0166 - accuracy: 0.9950 - val_loss: 1.4917 - val_accuracy: 0.7318\n",
      "Epoch 47/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0139 - accuracy: 0.9961 - val_loss: 1.6519 - val_accuracy: 0.7113\n",
      "Epoch 48/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0135 - accuracy: 0.9964 - val_loss: 1.5274 - val_accuracy: 0.7299\n",
      "Epoch 49/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0119 - accuracy: 0.9966 - val_loss: 1.5256 - val_accuracy: 0.7289\n",
      "Epoch 50/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0135 - accuracy: 0.9961 - val_loss: 1.5469 - val_accuracy: 0.7333\n",
      "Epoch 51/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0133 - accuracy: 0.9960 - val_loss: 1.5409 - val_accuracy: 0.7318\n",
      "Epoch 52/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0128 - accuracy: 0.9960 - val_loss: 1.5487 - val_accuracy: 0.7245\n",
      "Epoch 53/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0130 - accuracy: 0.9961 - val_loss: 1.5711 - val_accuracy: 0.7311\n",
      "Epoch 54/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0132 - accuracy: 0.9959 - val_loss: 1.5254 - val_accuracy: 0.7286\n",
      "Epoch 55/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0112 - accuracy: 0.9970 - val_loss: 1.5556 - val_accuracy: 0.7303\n",
      "Epoch 56/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0113 - accuracy: 0.9966 - val_loss: 1.5467 - val_accuracy: 0.7326\n",
      "Epoch 57/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0120 - accuracy: 0.9963 - val_loss: 1.5582 - val_accuracy: 0.7306\n",
      "Epoch 58/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0105 - accuracy: 0.9972 - val_loss: 1.5873 - val_accuracy: 0.7305\n",
      "Epoch 59/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0104 - accuracy: 0.9968 - val_loss: 1.5618 - val_accuracy: 0.7269\n",
      "Epoch 60/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0098 - accuracy: 0.9974 - val_loss: 1.5755 - val_accuracy: 0.7266\n",
      "Epoch 61/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0106 - accuracy: 0.9973 - val_loss: 1.6277 - val_accuracy: 0.7279\n",
      "Epoch 62/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0105 - accuracy: 0.9969 - val_loss: 1.5573 - val_accuracy: 0.7308\n",
      "Epoch 63/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0114 - accuracy: 0.9965 - val_loss: 1.5753 - val_accuracy: 0.7334\n",
      "Epoch 64/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0088 - accuracy: 0.9973 - val_loss: 1.6161 - val_accuracy: 0.7342\n",
      "Epoch 65/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0107 - accuracy: 0.9967 - val_loss: 1.5915 - val_accuracy: 0.7297\n",
      "Epoch 66/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0083 - accuracy: 0.9975 - val_loss: 1.5963 - val_accuracy: 0.7297\n",
      "Epoch 67/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0086 - accuracy: 0.9978 - val_loss: 1.6022 - val_accuracy: 0.7289\n",
      "Epoch 68/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0088 - accuracy: 0.9974 - val_loss: 1.6314 - val_accuracy: 0.7299\n",
      "Epoch 69/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0101 - accuracy: 0.9969 - val_loss: 1.6154 - val_accuracy: 0.7281\n",
      "Epoch 70/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0095 - accuracy: 0.9972 - val_loss: 1.6002 - val_accuracy: 0.7311\n",
      "Epoch 71/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0071 - accuracy: 0.9982 - val_loss: 1.6012 - val_accuracy: 0.7325\n",
      "Epoch 72/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0080 - accuracy: 0.9977 - val_loss: 1.6278 - val_accuracy: 0.7264\n",
      "Epoch 73/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.0081 - accuracy: 0.9980 - val_loss: 1.6060 - val_accuracy: 0.7359\n",
      "Epoch 74/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0085 - accuracy: 0.9974 - val_loss: 1.5912 - val_accuracy: 0.7346\n",
      "Epoch 75/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0074 - accuracy: 0.9978 - val_loss: 1.5889 - val_accuracy: 0.7312\n",
      "Epoch 76/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0081 - accuracy: 0.9974 - val_loss: 1.5992 - val_accuracy: 0.7386\n",
      "Epoch 77/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0071 - accuracy: 0.9981 - val_loss: 1.6418 - val_accuracy: 0.7318\n",
      "Epoch 78/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0070 - accuracy: 0.9979 - val_loss: 1.6251 - val_accuracy: 0.7310\n",
      "Epoch 79/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0081 - accuracy: 0.9976 - val_loss: 1.6070 - val_accuracy: 0.7341\n",
      "Epoch 80/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0070 - accuracy: 0.9980 - val_loss: 1.6386 - val_accuracy: 0.7370\n",
      "Epoch 81/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0062 - accuracy: 0.9983 - val_loss: 1.6504 - val_accuracy: 0.7300\n",
      "Epoch 82/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0078 - accuracy: 0.9975 - val_loss: 1.6405 - val_accuracy: 0.7380\n",
      "Epoch 83/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0068 - accuracy: 0.9978 - val_loss: 1.6144 - val_accuracy: 0.7372\n",
      "Epoch 84/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0064 - accuracy: 0.9983 - val_loss: 1.6274 - val_accuracy: 0.7356\n",
      "Epoch 85/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0070 - accuracy: 0.9980 - val_loss: 1.6258 - val_accuracy: 0.7357\n",
      "Epoch 86/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0059 - accuracy: 0.9987 - val_loss: 1.6383 - val_accuracy: 0.7318\n",
      "Epoch 87/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0066 - accuracy: 0.9981 - val_loss: 1.6337 - val_accuracy: 0.7333\n",
      "Epoch 88/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0066 - accuracy: 0.9982 - val_loss: 1.6579 - val_accuracy: 0.7355\n",
      "Epoch 89/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0069 - accuracy: 0.9982 - val_loss: 1.6468 - val_accuracy: 0.7324\n",
      "Epoch 90/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0070 - accuracy: 0.9981 - val_loss: 1.6400 - val_accuracy: 0.7354\n",
      "Epoch 91/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0060 - accuracy: 0.9985 - val_loss: 1.6457 - val_accuracy: 0.7336\n",
      "Epoch 92/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0059 - accuracy: 0.9983 - val_loss: 1.6435 - val_accuracy: 0.7372\n",
      "Epoch 93/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0061 - accuracy: 0.9983 - val_loss: 1.6792 - val_accuracy: 0.7365\n",
      "Epoch 94/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0050 - accuracy: 0.9988 - val_loss: 1.6552 - val_accuracy: 0.7354\n",
      "Epoch 95/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0055 - accuracy: 0.9984 - val_loss: 1.6751 - val_accuracy: 0.7353\n",
      "Epoch 96/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0057 - accuracy: 0.9986 - val_loss: 1.7063 - val_accuracy: 0.7283\n",
      "Epoch 97/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0059 - accuracy: 0.9982 - val_loss: 1.6727 - val_accuracy: 0.7358\n",
      "Epoch 98/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0045 - accuracy: 0.9990 - val_loss: 1.6578 - val_accuracy: 0.7374\n",
      "Epoch 99/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0052 - accuracy: 0.9988 - val_loss: 1.7025 - val_accuracy: 0.7328\n",
      "Epoch 100/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.0054 - accuracy: 0.9988 - val_loss: 1.6869 - val_accuracy: 0.7346\n",
      "Model saved as cifar_bn.h5\n",
      "History saved as hist_bn.p\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 30, 30, 32)        896       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 15, 15, 32)        128       \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 13, 13, 32)        9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 1152)              4608      \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 512)               590336    \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 512)               2048      \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 612,394\n",
      "Trainable params: 609,002\n",
      "Non-trainable params: 3,392\n",
      "_________________________________________________________________\n",
      "None\n",
      "Epoch 1/100\n",
      "1563/1563 [==============================] - 10s 6ms/step - loss: 2.0522 - accuracy: 0.3635 - val_loss: 1.3976 - val_accuracy: 0.5079\n",
      "Epoch 2/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 1.2912 - accuracy: 0.5398 - val_loss: 1.0957 - val_accuracy: 0.6072\n",
      "Epoch 3/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 1.1469 - accuracy: 0.5911 - val_loss: 1.0496 - val_accuracy: 0.6296\n",
      "Epoch 4/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 1.0714 - accuracy: 0.6187 - val_loss: 1.1098 - val_accuracy: 0.6102\n",
      "Epoch 5/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 1.0210 - accuracy: 0.6401 - val_loss: 0.8865 - val_accuracy: 0.6907\n",
      "Epoch 6/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.9609 - accuracy: 0.6609 - val_loss: 0.8370 - val_accuracy: 0.7037\n",
      "Epoch 7/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.9336 - accuracy: 0.6704 - val_loss: 0.8202 - val_accuracy: 0.7120\n",
      "Epoch 8/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.9099 - accuracy: 0.6783 - val_loss: 0.9419 - val_accuracy: 0.6727\n",
      "Epoch 9/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.8857 - accuracy: 0.6863 - val_loss: 0.7577 - val_accuracy: 0.7352\n",
      "Epoch 10/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.8611 - accuracy: 0.6989 - val_loss: 0.8067 - val_accuracy: 0.7186\n",
      "Epoch 11/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.8452 - accuracy: 0.7008 - val_loss: 0.7488 - val_accuracy: 0.7405\n",
      "Epoch 12/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.8231 - accuracy: 0.7115 - val_loss: 0.7977 - val_accuracy: 0.7198\n",
      "Epoch 13/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.8128 - accuracy: 0.7152 - val_loss: 0.7771 - val_accuracy: 0.7287\n",
      "Epoch 14/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.8063 - accuracy: 0.7152 - val_loss: 0.7384 - val_accuracy: 0.7421\n",
      "Epoch 15/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7948 - accuracy: 0.7206 - val_loss: 0.7828 - val_accuracy: 0.7259\n",
      "Epoch 16/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7775 - accuracy: 0.7266 - val_loss: 0.6956 - val_accuracy: 0.7574\n",
      "Epoch 17/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7636 - accuracy: 0.7312 - val_loss: 0.7022 - val_accuracy: 0.7568\n",
      "Epoch 18/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7583 - accuracy: 0.7315 - val_loss: 0.7575 - val_accuracy: 0.7363\n",
      "Epoch 19/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7553 - accuracy: 0.7363 - val_loss: 0.7350 - val_accuracy: 0.7469\n",
      "Epoch 20/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.7431 - accuracy: 0.7401 - val_loss: 0.7044 - val_accuracy: 0.7585\n",
      "Epoch 21/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7389 - accuracy: 0.7391 - val_loss: 0.6908 - val_accuracy: 0.7625\n",
      "Epoch 22/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7263 - accuracy: 0.7478 - val_loss: 0.6830 - val_accuracy: 0.7643\n",
      "Epoch 23/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.7246 - accuracy: 0.7436 - val_loss: 0.6546 - val_accuracy: 0.7761\n",
      "Epoch 24/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7208 - accuracy: 0.7489 - val_loss: 0.6604 - val_accuracy: 0.7782\n",
      "Epoch 25/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7129 - accuracy: 0.7484 - val_loss: 0.6904 - val_accuracy: 0.7587\n",
      "Epoch 26/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.7077 - accuracy: 0.7502 - val_loss: 0.6835 - val_accuracy: 0.7605\n",
      "Epoch 27/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6994 - accuracy: 0.7529 - val_loss: 0.7422 - val_accuracy: 0.7426\n",
      "Epoch 28/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6986 - accuracy: 0.7529 - val_loss: 0.7005 - val_accuracy: 0.7572\n",
      "Epoch 29/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6886 - accuracy: 0.7580 - val_loss: 0.6637 - val_accuracy: 0.7716\n",
      "Epoch 30/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6800 - accuracy: 0.7609 - val_loss: 0.7059 - val_accuracy: 0.7586\n",
      "Epoch 31/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6895 - accuracy: 0.7561 - val_loss: 0.6533 - val_accuracy: 0.7767\n",
      "Epoch 32/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6838 - accuracy: 0.7606 - val_loss: 0.6453 - val_accuracy: 0.7799\n",
      "Epoch 33/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6717 - accuracy: 0.7686 - val_loss: 0.6437 - val_accuracy: 0.7802\n",
      "Epoch 34/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6797 - accuracy: 0.7628 - val_loss: 0.6491 - val_accuracy: 0.7788\n",
      "Epoch 35/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6669 - accuracy: 0.7635 - val_loss: 0.6850 - val_accuracy: 0.7674\n",
      "Epoch 36/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6709 - accuracy: 0.7640 - val_loss: 0.6553 - val_accuracy: 0.7724\n",
      "Epoch 37/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6616 - accuracy: 0.7660 - val_loss: 0.6498 - val_accuracy: 0.7829\n",
      "Epoch 38/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6668 - accuracy: 0.7654 - val_loss: 0.6851 - val_accuracy: 0.7614\n",
      "Epoch 39/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.6582 - accuracy: 0.7729 - val_loss: 0.6219 - val_accuracy: 0.7891\n",
      "Epoch 40/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6500 - accuracy: 0.7711 - val_loss: 0.6296 - val_accuracy: 0.7884\n",
      "Epoch 41/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6545 - accuracy: 0.7685 - val_loss: 0.6242 - val_accuracy: 0.7870\n",
      "Epoch 42/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6473 - accuracy: 0.7683 - val_loss: 0.6250 - val_accuracy: 0.7852\n",
      "Epoch 43/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6472 - accuracy: 0.7714 - val_loss: 0.6532 - val_accuracy: 0.7775\n",
      "Epoch 44/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6410 - accuracy: 0.7729 - val_loss: 0.6175 - val_accuracy: 0.7919\n",
      "Epoch 45/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6475 - accuracy: 0.7752 - val_loss: 0.6367 - val_accuracy: 0.7825\n",
      "Epoch 46/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6420 - accuracy: 0.7712 - val_loss: 0.6552 - val_accuracy: 0.7747\n",
      "Epoch 47/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6358 - accuracy: 0.7775 - val_loss: 0.6326 - val_accuracy: 0.7868\n",
      "Epoch 48/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6415 - accuracy: 0.7751 - val_loss: 0.6437 - val_accuracy: 0.7782\n",
      "Epoch 49/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6355 - accuracy: 0.7762 - val_loss: 0.6181 - val_accuracy: 0.7913\n",
      "Epoch 50/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6217 - accuracy: 0.7798 - val_loss: 0.6506 - val_accuracy: 0.7828\n",
      "Epoch 51/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6175 - accuracy: 0.7819 - val_loss: 0.6213 - val_accuracy: 0.7890\n",
      "Epoch 52/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6208 - accuracy: 0.7861 - val_loss: 0.6310 - val_accuracy: 0.7852\n",
      "Epoch 53/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6231 - accuracy: 0.7795 - val_loss: 0.6682 - val_accuracy: 0.7696\n",
      "Epoch 54/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6239 - accuracy: 0.7802 - val_loss: 0.6371 - val_accuracy: 0.7846\n",
      "Epoch 55/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6094 - accuracy: 0.7825 - val_loss: 0.6084 - val_accuracy: 0.7933\n",
      "Epoch 56/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6108 - accuracy: 0.7834 - val_loss: 0.6135 - val_accuracy: 0.7967\n",
      "Epoch 57/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6148 - accuracy: 0.7810 - val_loss: 0.6307 - val_accuracy: 0.7837\n",
      "Epoch 58/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6002 - accuracy: 0.7853 - val_loss: 0.6148 - val_accuracy: 0.7915\n",
      "Epoch 59/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6223 - accuracy: 0.7793 - val_loss: 0.6122 - val_accuracy: 0.7941\n",
      "Epoch 60/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6158 - accuracy: 0.7808 - val_loss: 0.6199 - val_accuracy: 0.7873\n",
      "Epoch 61/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6073 - accuracy: 0.7828 - val_loss: 0.6081 - val_accuracy: 0.7974\n",
      "Epoch 62/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6003 - accuracy: 0.7874 - val_loss: 0.6053 - val_accuracy: 0.7978\n",
      "Epoch 63/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6055 - accuracy: 0.7859 - val_loss: 0.6102 - val_accuracy: 0.7924\n",
      "Epoch 64/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6020 - accuracy: 0.7900 - val_loss: 0.6086 - val_accuracy: 0.7969\n",
      "Epoch 65/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.6080 - accuracy: 0.7845 - val_loss: 0.6098 - val_accuracy: 0.7945\n",
      "Epoch 66/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5992 - accuracy: 0.7899 - val_loss: 0.6279 - val_accuracy: 0.7873\n",
      "Epoch 67/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6005 - accuracy: 0.7884 - val_loss: 0.6082 - val_accuracy: 0.7975\n",
      "Epoch 68/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5926 - accuracy: 0.7935 - val_loss: 0.6630 - val_accuracy: 0.7716\n",
      "Epoch 69/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6040 - accuracy: 0.7886 - val_loss: 0.6314 - val_accuracy: 0.7820\n",
      "Epoch 70/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.6009 - accuracy: 0.7890 - val_loss: 0.6132 - val_accuracy: 0.7932\n",
      "Epoch 71/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.6052 - accuracy: 0.7868 - val_loss: 0.6102 - val_accuracy: 0.7916\n",
      "Epoch 72/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5907 - accuracy: 0.7910 - val_loss: 0.5986 - val_accuracy: 0.7976\n",
      "Epoch 73/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5876 - accuracy: 0.7921 - val_loss: 0.6011 - val_accuracy: 0.7971\n",
      "Epoch 74/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5921 - accuracy: 0.7911 - val_loss: 0.6124 - val_accuracy: 0.7914\n",
      "Epoch 75/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5839 - accuracy: 0.7969 - val_loss: 0.5981 - val_accuracy: 0.7973\n",
      "Epoch 76/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5893 - accuracy: 0.7917 - val_loss: 0.6161 - val_accuracy: 0.7906\n",
      "Epoch 77/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5871 - accuracy: 0.7918 - val_loss: 0.6012 - val_accuracy: 0.7986\n",
      "Epoch 78/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5790 - accuracy: 0.7945 - val_loss: 0.6013 - val_accuracy: 0.7950\n",
      "Epoch 79/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5819 - accuracy: 0.7961 - val_loss: 0.6221 - val_accuracy: 0.7912\n",
      "Epoch 80/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5960 - accuracy: 0.7891 - val_loss: 0.6042 - val_accuracy: 0.7932\n",
      "Epoch 81/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5902 - accuracy: 0.7909 - val_loss: 0.6054 - val_accuracy: 0.7954\n",
      "Epoch 82/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5780 - accuracy: 0.7967 - val_loss: 0.6049 - val_accuracy: 0.7941\n",
      "Epoch 83/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5771 - accuracy: 0.7953 - val_loss: 0.6120 - val_accuracy: 0.7904\n",
      "Epoch 84/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5667 - accuracy: 0.8001 - val_loss: 0.6122 - val_accuracy: 0.7962\n",
      "Epoch 85/100\n",
      "1563/1563 [==============================] - 9s 5ms/step - loss: 0.5708 - accuracy: 0.7989 - val_loss: 0.6001 - val_accuracy: 0.7953\n",
      "Epoch 86/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5732 - accuracy: 0.7991 - val_loss: 0.6066 - val_accuracy: 0.7939\n",
      "Epoch 87/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5693 - accuracy: 0.7985 - val_loss: 0.5959 - val_accuracy: 0.7985\n",
      "Epoch 88/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5797 - accuracy: 0.7967 - val_loss: 0.5930 - val_accuracy: 0.7984\n",
      "Epoch 89/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5848 - accuracy: 0.7925 - val_loss: 0.6270 - val_accuracy: 0.7837\n",
      "Epoch 90/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5641 - accuracy: 0.7985 - val_loss: 0.6078 - val_accuracy: 0.7943\n",
      "Epoch 91/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5866 - accuracy: 0.7941 - val_loss: 0.5898 - val_accuracy: 0.8017\n",
      "Epoch 92/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5758 - accuracy: 0.7986 - val_loss: 0.5980 - val_accuracy: 0.7978\n",
      "Epoch 93/100\n",
      "1563/1563 [==============================] - 8s 5ms/step - loss: 0.5696 - accuracy: 0.7976 - val_loss: 0.6102 - val_accuracy: 0.7922\n",
      "Epoch 94/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5779 - accuracy: 0.7964 - val_loss: 0.6005 - val_accuracy: 0.7941\n",
      "Epoch 95/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5707 - accuracy: 0.7968 - val_loss: 0.5924 - val_accuracy: 0.7977\n",
      "Epoch 96/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5597 - accuracy: 0.7980 - val_loss: 0.6196 - val_accuracy: 0.7874\n",
      "Epoch 97/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5615 - accuracy: 0.8025 - val_loss: 0.6065 - val_accuracy: 0.7953\n",
      "Epoch 98/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5696 - accuracy: 0.7993 - val_loss: 0.6070 - val_accuracy: 0.7919\n",
      "Epoch 99/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5677 - accuracy: 0.7984 - val_loss: 0.6093 - val_accuracy: 0.7959\n",
      "Epoch 100/100\n",
      "1563/1563 [==============================] - 9s 6ms/step - loss: 0.5694 - accuracy: 0.7981 - val_loss: 0.6023 - val_accuracy: 0.7969\n",
      "Model saved as cifar_dropout.h5\n",
      "History saved as hist_dropout.p\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 30, 30, 32)        896       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 15, 15, 32)        128       \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 13, 13, 32)        9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 1152)              4608      \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1152)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 512)               590336    \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 512)               2048      \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 612,394\n",
      "Trainable params: 609,002\n",
      "Non-trainable params: 3,392\n",
      "_________________________________________________________________\n",
      "None\n",
      "Epoch 1/100\n",
      "1563/1563 [==============================] - 32s 20ms/step - loss: 2.1087 - accuracy: 0.3447 - val_loss: 1.2918 - val_accuracy: 0.5353\n",
      "Epoch 2/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.4204 - accuracy: 0.4922 - val_loss: 1.0644 - val_accuracy: 0.6217\n",
      "Epoch 3/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.3083 - accuracy: 0.5307 - val_loss: 1.2747 - val_accuracy: 0.5386\n",
      "Epoch 4/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.2342 - accuracy: 0.5606 - val_loss: 1.0155 - val_accuracy: 0.6505\n",
      "Epoch 5/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.2022 - accuracy: 0.5757 - val_loss: 1.0120 - val_accuracy: 0.6351\n",
      "Epoch 6/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.1583 - accuracy: 0.5869 - val_loss: 0.9384 - val_accuracy: 0.6675\n",
      "Epoch 7/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.1255 - accuracy: 0.6011 - val_loss: 0.8839 - val_accuracy: 0.6928\n",
      "Epoch 8/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.1005 - accuracy: 0.6116 - val_loss: 0.9405 - val_accuracy: 0.6662\n",
      "Epoch 9/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0850 - accuracy: 0.6158 - val_loss: 0.9075 - val_accuracy: 0.6715\n",
      "Epoch 10/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0589 - accuracy: 0.6279 - val_loss: 0.8753 - val_accuracy: 0.6981\n",
      "Epoch 11/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0460 - accuracy: 0.6315 - val_loss: 0.8634 - val_accuracy: 0.7010\n",
      "Epoch 12/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0369 - accuracy: 0.6340 - val_loss: 0.8409 - val_accuracy: 0.7022\n",
      "Epoch 13/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0190 - accuracy: 0.6418 - val_loss: 0.8533 - val_accuracy: 0.7062\n",
      "Epoch 14/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0189 - accuracy: 0.6359 - val_loss: 0.8472 - val_accuracy: 0.7028\n",
      "Epoch 15/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 1.0085 - accuracy: 0.6421 - val_loss: 0.7731 - val_accuracy: 0.7247\n",
      "Epoch 16/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9982 - accuracy: 0.6461 - val_loss: 0.7770 - val_accuracy: 0.7311\n",
      "Epoch 17/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9829 - accuracy: 0.6536 - val_loss: 0.8865 - val_accuracy: 0.6880\n",
      "Epoch 18/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9799 - accuracy: 0.6547 - val_loss: 0.8172 - val_accuracy: 0.7134\n",
      "Epoch 19/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9713 - accuracy: 0.6588 - val_loss: 0.7613 - val_accuracy: 0.7358\n",
      "Epoch 20/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9689 - accuracy: 0.6589 - val_loss: 0.8063 - val_accuracy: 0.7183\n",
      "Epoch 21/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9703 - accuracy: 0.6619 - val_loss: 0.8115 - val_accuracy: 0.7165\n",
      "Epoch 22/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9649 - accuracy: 0.6612 - val_loss: 0.7411 - val_accuracy: 0.7436\n",
      "Epoch 23/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9545 - accuracy: 0.6680 - val_loss: 0.8145 - val_accuracy: 0.7165\n",
      "Epoch 24/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9482 - accuracy: 0.6678 - val_loss: 0.8218 - val_accuracy: 0.7172\n",
      "Epoch 25/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9520 - accuracy: 0.6683 - val_loss: 0.7904 - val_accuracy: 0.7247\n",
      "Epoch 26/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9520 - accuracy: 0.6655 - val_loss: 0.7319 - val_accuracy: 0.7504\n",
      "Epoch 27/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9475 - accuracy: 0.6668 - val_loss: 0.7101 - val_accuracy: 0.7520\n",
      "Epoch 28/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9357 - accuracy: 0.6722 - val_loss: 0.7225 - val_accuracy: 0.7517\n",
      "Epoch 29/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9340 - accuracy: 0.6720 - val_loss: 0.7317 - val_accuracy: 0.7470\n",
      "Epoch 30/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9331 - accuracy: 0.6737 - val_loss: 0.6967 - val_accuracy: 0.7616\n",
      "Epoch 31/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9250 - accuracy: 0.6743 - val_loss: 0.7102 - val_accuracy: 0.7566\n",
      "Epoch 32/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9222 - accuracy: 0.6786 - val_loss: 0.7061 - val_accuracy: 0.7550\n",
      "Epoch 33/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9203 - accuracy: 0.6744 - val_loss: 0.7054 - val_accuracy: 0.7554\n",
      "Epoch 34/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9280 - accuracy: 0.6738 - val_loss: 0.7110 - val_accuracy: 0.7533\n",
      "Epoch 35/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9045 - accuracy: 0.6824 - val_loss: 0.7185 - val_accuracy: 0.7509\n",
      "Epoch 36/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9108 - accuracy: 0.6791 - val_loss: 0.7545 - val_accuracy: 0.7367\n",
      "Epoch 37/100\n",
      "1563/1563 [==============================] - 32s 20ms/step - loss: 0.9015 - accuracy: 0.6847 - val_loss: 0.6976 - val_accuracy: 0.7571\n",
      "Epoch 38/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9202 - accuracy: 0.6790 - val_loss: 0.6972 - val_accuracy: 0.7587\n",
      "Epoch 39/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9021 - accuracy: 0.6835 - val_loss: 0.7242 - val_accuracy: 0.7449\n",
      "Epoch 40/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9013 - accuracy: 0.6871 - val_loss: 0.7324 - val_accuracy: 0.7489\n",
      "Epoch 41/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9024 - accuracy: 0.6831 - val_loss: 0.7055 - val_accuracy: 0.7572\n",
      "Epoch 42/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.9065 - accuracy: 0.6831 - val_loss: 0.7054 - val_accuracy: 0.7539\n",
      "Epoch 43/100\n",
      "1563/1563 [==============================] - 32s 20ms/step - loss: 0.9013 - accuracy: 0.6839 - val_loss: 0.6979 - val_accuracy: 0.7569\n",
      "Epoch 44/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8861 - accuracy: 0.6928 - val_loss: 0.6868 - val_accuracy: 0.7622\n",
      "Epoch 45/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8884 - accuracy: 0.6866 - val_loss: 0.7142 - val_accuracy: 0.7584\n",
      "Epoch 46/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8941 - accuracy: 0.6885 - val_loss: 0.7040 - val_accuracy: 0.7577\n",
      "Epoch 47/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8803 - accuracy: 0.6931 - val_loss: 0.7010 - val_accuracy: 0.7545\n",
      "Epoch 48/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8930 - accuracy: 0.6887 - val_loss: 0.7064 - val_accuracy: 0.7577\n",
      "Epoch 49/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8820 - accuracy: 0.6923 - val_loss: 0.7310 - val_accuracy: 0.7450\n",
      "Epoch 50/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8785 - accuracy: 0.6920 - val_loss: 0.6787 - val_accuracy: 0.7674\n",
      "Epoch 51/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8805 - accuracy: 0.6968 - val_loss: 0.6705 - val_accuracy: 0.7702\n",
      "Epoch 52/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8846 - accuracy: 0.6901 - val_loss: 0.6824 - val_accuracy: 0.7641\n",
      "Epoch 53/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8850 - accuracy: 0.6913 - val_loss: 0.6925 - val_accuracy: 0.7566\n",
      "Epoch 54/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8719 - accuracy: 0.6942 - val_loss: 0.6675 - val_accuracy: 0.7697\n",
      "Epoch 55/100\n",
      "1563/1563 [==============================] - 32s 20ms/step - loss: 0.8690 - accuracy: 0.6981 - val_loss: 0.6699 - val_accuracy: 0.7685\n",
      "Epoch 56/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8738 - accuracy: 0.6951 - val_loss: 0.6725 - val_accuracy: 0.7669\n",
      "Epoch 57/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8687 - accuracy: 0.6947 - val_loss: 0.7001 - val_accuracy: 0.7552\n",
      "Epoch 58/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8709 - accuracy: 0.6952 - val_loss: 0.6653 - val_accuracy: 0.7692\n",
      "Epoch 59/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8731 - accuracy: 0.6940 - val_loss: 0.6615 - val_accuracy: 0.7709\n",
      "Epoch 60/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8616 - accuracy: 0.6979 - val_loss: 0.7046 - val_accuracy: 0.7503\n",
      "Epoch 61/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8600 - accuracy: 0.6976 - val_loss: 0.6612 - val_accuracy: 0.7704\n",
      "Epoch 62/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8647 - accuracy: 0.7012 - val_loss: 0.6650 - val_accuracy: 0.7698\n",
      "Epoch 63/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8789 - accuracy: 0.6952 - val_loss: 0.6620 - val_accuracy: 0.7703\n",
      "Epoch 64/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8717 - accuracy: 0.6968 - val_loss: 0.6590 - val_accuracy: 0.7744\n",
      "Epoch 65/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8495 - accuracy: 0.7035 - val_loss: 0.6674 - val_accuracy: 0.7666\n",
      "Epoch 66/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8628 - accuracy: 0.6985 - val_loss: 0.6484 - val_accuracy: 0.7741\n",
      "Epoch 67/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8700 - accuracy: 0.6943 - val_loss: 0.6607 - val_accuracy: 0.7707\n",
      "Epoch 68/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8551 - accuracy: 0.6995 - val_loss: 0.6412 - val_accuracy: 0.7767\n",
      "Epoch 69/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8514 - accuracy: 0.7017 - val_loss: 0.6536 - val_accuracy: 0.7775\n",
      "Epoch 70/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8646 - accuracy: 0.6976 - val_loss: 0.6608 - val_accuracy: 0.7700\n",
      "Epoch 71/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8546 - accuracy: 0.7023 - val_loss: 0.6776 - val_accuracy: 0.7643\n",
      "Epoch 72/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8642 - accuracy: 0.6987 - val_loss: 0.6675 - val_accuracy: 0.7667\n",
      "Epoch 73/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8473 - accuracy: 0.7038 - val_loss: 0.6424 - val_accuracy: 0.7779\n",
      "Epoch 74/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8527 - accuracy: 0.7012 - val_loss: 0.6713 - val_accuracy: 0.7694\n",
      "Epoch 75/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8608 - accuracy: 0.6977 - val_loss: 0.6469 - val_accuracy: 0.7757\n",
      "Epoch 76/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8512 - accuracy: 0.7026 - val_loss: 0.6415 - val_accuracy: 0.7767\n",
      "Epoch 77/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8510 - accuracy: 0.7012 - val_loss: 0.6426 - val_accuracy: 0.7766\n",
      "Epoch 78/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8453 - accuracy: 0.7067 - val_loss: 0.6480 - val_accuracy: 0.7750\n",
      "Epoch 79/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8575 - accuracy: 0.7009 - val_loss: 0.6399 - val_accuracy: 0.7776\n",
      "Epoch 80/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8481 - accuracy: 0.7035 - val_loss: 0.6512 - val_accuracy: 0.7758\n",
      "Epoch 81/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8500 - accuracy: 0.7036 - val_loss: 0.6431 - val_accuracy: 0.7800\n",
      "Epoch 82/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8548 - accuracy: 0.7029 - val_loss: 0.6466 - val_accuracy: 0.7762\n",
      "Epoch 83/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8380 - accuracy: 0.7058 - val_loss: 0.6552 - val_accuracy: 0.7723\n",
      "Epoch 84/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8554 - accuracy: 0.7011 - val_loss: 0.6506 - val_accuracy: 0.7741\n",
      "Epoch 85/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8472 - accuracy: 0.7079 - val_loss: 0.6464 - val_accuracy: 0.7758\n",
      "Epoch 86/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8440 - accuracy: 0.7067 - val_loss: 0.6322 - val_accuracy: 0.7799\n",
      "Epoch 87/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8483 - accuracy: 0.7024 - val_loss: 0.6393 - val_accuracy: 0.7782\n",
      "Epoch 88/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8479 - accuracy: 0.7064 - val_loss: 0.6463 - val_accuracy: 0.7735\n",
      "Epoch 89/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8491 - accuracy: 0.7024 - val_loss: 0.6302 - val_accuracy: 0.7803\n",
      "Epoch 90/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8412 - accuracy: 0.7061 - val_loss: 0.6476 - val_accuracy: 0.7770\n",
      "Epoch 91/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8432 - accuracy: 0.7047 - val_loss: 0.6386 - val_accuracy: 0.7780\n",
      "Epoch 92/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8492 - accuracy: 0.7047 - val_loss: 0.6715 - val_accuracy: 0.7643\n",
      "Epoch 93/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8382 - accuracy: 0.7085 - val_loss: 0.6448 - val_accuracy: 0.7774\n",
      "Epoch 94/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8417 - accuracy: 0.7060 - val_loss: 0.6295 - val_accuracy: 0.7802\n",
      "Epoch 95/100\n",
      "1563/1563 [==============================] - 32s 20ms/step - loss: 0.8309 - accuracy: 0.7089 - val_loss: 0.6598 - val_accuracy: 0.7676\n",
      "Epoch 96/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8282 - accuracy: 0.7115 - val_loss: 0.6314 - val_accuracy: 0.7808\n",
      "Epoch 97/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8458 - accuracy: 0.7021 - val_loss: 0.6502 - val_accuracy: 0.7731\n",
      "Epoch 98/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8391 - accuracy: 0.7049 - val_loss: 0.6324 - val_accuracy: 0.7809\n",
      "Epoch 99/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8281 - accuracy: 0.7087 - val_loss: 0.6408 - val_accuracy: 0.7763\n",
      "Epoch 100/100\n",
      "1563/1563 [==============================] - 31s 20ms/step - loss: 0.8321 - accuracy: 0.7062 - val_loss: 0.6386 - val_accuracy: 0.7773\n",
      "Model saved as cifar_dataaug.h5\n",
      "History saved as hist_dataaug.p\n"
     ]
    }
   ],
   "source": [
    "mod_name_test = ['basic', 'bn', 'dropout', 'dataaug']\n",
    "nmeth = len(mod_name_test)\n",
    "\n",
    "# Parameters\n",
    "nepochs = 100\n",
    "batch_size = 32\n",
    "lr = 1e-3\n",
    "decay = 1e-4\n",
    "\n",
    "# Main loop\n",
    "for i, mod_name in enumerate(mod_name_test):\n",
    "\n",
    "    # Create the model based on the model type\n",
    "    K.clear_session()\n",
    "    if mod_name == 'basic':\n",
    "        model = create_mod()\n",
    "    elif mod_name == 'bn':\n",
    "        model = create_mod(use_bn=True)\n",
    "    elif (mod_name == 'dropout') or (mod_name == 'dataaug'):\n",
    "        model = create_mod(use_bn=True, use_dropout=True)        \n",
    "    else:\n",
    "        raise Exception('Model %s not known' % mod_name)\n",
    "\n",
    "    # Create the optimizer\n",
    "    opt = optimizers.Adam(lr=lr, decay=decay)\n",
    "\n",
    "    # Compile\n",
    "    hist = model.compile(loss='categorical_crossentropy',\n",
    "                  optimizer=opt,\n",
    "                  metrics=['accuracy'])\n",
    "    print(model.summary()) \n",
    "\n",
    "    if mod_name == 'dataaug':\n",
    "\n",
    "        \n",
    "        \n",
    "        # Create a data augmentation \n",
    "        datagen = create_datagen()\n",
    "\n",
    "        # Compute quantities required for feature-wise normalization\n",
    "        # (std, mean, and principal components if ZCA whitening is applied).\n",
    "        datagen.fit(Xtr)\n",
    "\n",
    "        # Fit the model on the batches generated by datagen.flow().\n",
    "        hist = model.fit(datagen.flow(Xtr, ytr_cat,\n",
    "                            batch_size=batch_size),\n",
    "                            epochs=nepochs, validation_data=(Xts,yts_cat))\n",
    "\n",
    "    else:\n",
    "        # Fit the model with no data augmentation\n",
    "        hist = model.fit(Xtr, ytr_cat, batch_size=batch_size,\n",
    "              epochs=nepochs, validation_data=(Xts, yts_cat),\n",
    "              shuffle=True)\n",
    "\n",
    "    # Save model\n",
    "    h5_fn = ('cifar_%s.h5' % mod_name)\n",
    "    model.save(h5_fn)\n",
    "    print('Model saved as %s' % h5_fn)\n",
    "\n",
    "    # Save history\n",
    "    hist_fn = ('hist_%s.p' % mod_name)\n",
    "    with open(hist_fn, 'wb') as fp:\n",
    "        hist_dict = hist.history\n",
    "        pickle.dump(hist_dict, fp) \n",
    "    print('History saved as %s' % hist_fn)                                "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5EJ4d3j0HvIT"
   },
   "source": [
    "We now plot the results.  You should approximately get\n",
    "\n",
    "* `baseline` ~70%\n",
    "* `bn` ~74%\n",
    "* `bn+dropout` ~78%\n",
    "* `bn+dropout+augmentation` ~76%\n",
    "\n",
    "So batch normalization and dropout help, but augmentation is not necessary in this case.  Note that `baseline` and `bn` get training accuracies of ~100% suggesting overfitting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 369
    },
    "id": "kREhQ0v4HvIT",
    "outputId": "98b77532-a1ea-4b52-8931-2edd56236849"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mod_name_plot = ['basic', 'bn', 'dropout', 'dataaug']\n",
    "plt.figure(figsize=(10,5))\n",
    "for iplt in range(2):\n",
    "    \n",
    "    plt.subplot(1,2,iplt+1)\n",
    "    for i, mod_name in enumerate(mod_name_plot):\n",
    "\n",
    "        # Load history\n",
    "        hist_fn = ('hist_%s.p' % mod_name)\n",
    "        with open(hist_fn, 'rb') as fp:        \n",
    "            hist_dict = pickle.load(fp) \n",
    "\n",
    "        if iplt == 0:\n",
    "            acc = hist_dict['accuracy']\n",
    "        else:\n",
    "            acc = hist_dict['val_accuracy']\n",
    "        plt.plot(acc, '-', linewidth=3)\n",
    "    \n",
    "    n = len(acc)\n",
    "    nepochs = len(acc)\n",
    "    plt.grid()\n",
    "    plt.xlim([0, nepochs])\n",
    "    plt.legend(['baseline', 'bn', 'bn+dropout', 'bn+dropout+aug'])\n",
    "    plt.xlabel('Epoch')\n",
    "    if iplt == 0:\n",
    "        plt.ylabel('Train accuracy')\n",
    "    else:\n",
    "        plt.ylabel('Test accuracy')\n",
    "        \n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9WGmbKH-HvIT"
   },
   "source": [
    "Print final accuracies:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "lxkoMdTcHvIT",
    "outputId": "8837bd19-990d-4e76-bcad-031e4f11f71b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "          basic :  0.708\n",
      "             bn :  0.734\n",
      "        dropout :  0.793\n",
      "        dataaug :  0.778\n"
     ]
    }
   ],
   "source": [
    "for i, mod_name in enumerate(mod_name_plot):\n",
    "\n",
    "    # Load history\n",
    "    hist_fn = ('hist_%s.p' % mod_name)\n",
    "    with open(hist_fn, 'rb') as fp:        \n",
    "        hist_dict = pickle.load(fp) \n",
    "        \n",
    "    # Print average of last 5 acc\n",
    "    acc = hist_dict['val_accuracy']\n",
    "    print('%15s :  %5.3f' % (mod_name, np.mean(acc[-5:])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "CM-RapB7HvIT"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "machine_shape": "hm",
   "name": "demo4_classifier.ipynb",
   "provenance": []
  },
  "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}