codearch/train_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "SWg0x3YTr6hN",
    "outputId": "f1a80a9d-4aae-4317-c5bc-b9f8bed15f95"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "71iB3EZ5KZna",
    "outputId": "8b2f5dee-971e-45cc-a193-575035cfaaf1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 60,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%pylab inline\n",
    "import tensorflow as tf\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from tensorflow.keras.models import Sequential, load_model\n",
    "from tensorflow.keras.layers import AveragePooling2D, GlobalAveragePooling2D, Conv2D, MaxPooling2D, Dense, Activation, Dropout, Flatten, BatchNormalization, SeparableConv2D, Input\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.regularizers import l2\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.initializers import RandomNormal\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, TensorBoard, ReduceLROnPlateau\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras import layers\n",
    "import pickle\n",
    "\n",
    "# Checking if GPU is available\n",
    "tf.test.gpu_device_name()\n",
    "tf.test.is_gpu_available(cuda_only=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ZF1H41Fdr6hS"
   },
   "outputs": [],
   "source": [
    "X = pickle.load(open('/content/drive/My Drive/brain tumor/X.pickle','rb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vucC94Uor6hU"
   },
   "outputs": [],
   "source": [
    "y = np.load('/content/drive/My Drive/brain tumor/y.npy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "brEO3OC0hCPK",
    "outputId": "d672e8d0-f72d-41cb-d43e-f4f6f9589ccb"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:3: RuntimeWarning: invalid value encountered in true_divide\n",
      "  This is separate from the ipykernel package so we can avoid doing imports until\n"
     ]
    }
   ],
   "source": [
    "each_pixel_mean = X.mean(axis=0)\n",
    "each_pixel_std = np.std(X, axis=0)\n",
    "images = np.divide(np.subtract(X,each_pixel_mean), each_pixel_std)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "sEXnWKoTLXGS",
    "outputId": "770b723b-e5bc-43e1-a057-1029cccd7de7"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3064, 256, 256, 1)"
      ]
     },
     "execution_count": 13,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TafL0Gdjr6hX"
   },
   "outputs": [],
   "source": [
    "X = X.astype('float32')\n",
    "y = to_categorical(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KjNkcpIMr6ha"
   },
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, shuffle = False)\n",
    "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.5, shuffle = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "uLSa-ouPr6hc"
   },
   "outputs": [],
   "source": [
    "'''\n",
    "def architecture(input_shape, num_classes, l2_regularization=0.001):\n",
    "    regularization = l2(l2_regularization)\n",
    "\n",
    "    \n",
    "    img_input = Input(input_shape)\n",
    "    \n",
    "    #Base block\n",
    "    x = Conv2D(filters=64, kernel_size=(5, 5), strides=(1, 1), padding='same', activation='relu', kernel_regularizer=regularization, use_bias=False)(img_input)\n",
    "    residual_base = x\n",
    "    residual_base_128 = tf.reshape(residual_base, shape=[-1, 256, 256, 128])\n",
    "    residual_base_256 = tf.reshape(residual_base, shape=[-1, 256, 256, 256])\n",
    "    residual_base_512 = tf.reshape(residual_base, shape=[-1, 256, 256, 512])\n",
    "    \n",
    "    #First residual block\n",
    "    x = Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu', kernel_regularizer=regularization, use_bias=False)(x)\n",
    "    x = Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu', kernel_regularizer=regularization, use_bias=False)(x)\n",
    "    residual_1 = x \n",
    "    residual_1_128 = tf.reshape(residual_1, shape=[-1, 256, 256, 128])\n",
    "    residual_1_256 = tf.reshape(residual_1, shape=[-1, 256, 256, 256])\n",
    "    residual_1_512 = tf.reshape(residual_1, shape=[-1, 256, 256, 512])\n",
    "   \n",
    "    #Second residual block\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation='relu', padding='same', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation='relu', padding='same', use_bias=False)(x)\n",
    "    residual_2 = x\n",
    "    residual_2_256 = tf.reshape(residual_1, shape=[-1, 256, 256, 256])\n",
    "    residual_2_512 = tf.reshape(residual_1, shape=[-1, 256, 256, 512])\n",
    "\n",
    "    x = layers.add([x,residual_1_128])\n",
    "    \n",
    "\n",
    "    #Third residual block\n",
    "    x = Conv2D(filters=256, kernel_size=(3, 3), padding='same', activation='relu', use_bias=False)(x)\n",
    "    x = Conv2D(filters=256, kernel_size=(3, 3), padding='same',  activation='relu', use_bias=False)(x)\n",
    "    residual_3 = x\n",
    "    residual_3_512 = tf.reshape(residual_3, shape=[-1, 256,256,512])\n",
    "    x = layers.add([x, residual_2_256])\n",
    "    \n",
    "    #Fourth residual block\n",
    "    x = Conv2D(filters=512, kernel_size=(3, 3), padding='same',  activation='relu', use_bias=False)(x)\n",
    "    x = Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu', use_bias=False)(x)\n",
    "    x = layers.add([x, residual_3_512])\n",
    "    \n",
    "    x = GlobalAveragePooling2D()(x)\n",
    "   \n",
    "    \n",
    "    output = Dense(units=3, activation='softmax')(x)\n",
    "    \n",
    "\n",
    "    model = Model(img_input, output)\n",
    "    model.summary()\n",
    "    return model\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "p7UO-TJll-eo"
   },
   "outputs": [],
   "source": [
    "def architecture(input_shape, num_classes, l2_regularization=0.01):\n",
    "    regularization = l2(l2_regularization)\n",
    "\n",
    "    \n",
    "    img_input = Input(input_shape)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu', kernel_regularizer=regularization, use_bias=False)(img_input)\n",
    "    #x = MaxPooling2D(pool_size=(2, 2), padding='same')(x)\n",
    "    #x = Dropout(0.10)(x)\n",
    "    \n",
    "    residual_1 = MaxPooling2D()(x)\n",
    "\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation='relu', padding='same', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation='relu', padding='same', use_bias=False)(x)\n",
    "    #x = Dropout(0.10)(x)\n",
    "    x = MaxPooling2D()(x)\n",
    "    x = layers.add([x,residual_1])\n",
    "    \n",
    "    residual_2 = x\n",
    "    \n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    #x = Dropout(0.10)(x)\n",
    "    x = MaxPooling2D()(x)\n",
    "    x = layers.add([x, residual_1, residual_2])\n",
    "    \n",
    "    residual_3 = MaxPooling2D()(x)\n",
    "    \n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    #x = Dropout(0.10)(x)\n",
    "    x = MaxPooling2D()(x)\n",
    "    x = layers.add([x, residual_1, residual_2, residual_3])\n",
    "    \n",
    "    \n",
    "    x = Conv2D(filters=128, kernel_size=(3,3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = GlobalAveragePooling2D()(x)\n",
    "    #y = GlobalAveragePooling2D()(residual_1)\n",
    "    #z = GlobalAveragePooling2D()(residual_2)\n",
    "\n",
    "    #x_y_z = layers.add([x,y,z])\n",
    "    \n",
    "    \n",
    "    output = Dense(units=3, activation='softmax')(x)\n",
    "    \n",
    "\n",
    "    model = Model(img_input, output)\n",
    "    model.summary()\n",
    "    return model\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "NpDHS_kXiKNe"
   },
   "outputs": [],
   "source": [
    "def architecture(input_shape, num_classes, l2_regularization=0.01):\n",
    "    regularization = l2(l2_regularization)\n",
    "\n",
    "    \n",
    "    img_input = Input(input_shape)\n",
    "    x = Conv2D(filters=128, kernel_size=(5, 5), strides=(1, 1), padding='same', activation='relu', kernel_regularizer=regularization, use_bias=False)(img_input)\n",
    "    x = MaxPooling2D(pool_size=(2, 2), padding='same')(x)\n",
    "    x = Dropout(0.10)(x)\n",
    "    \n",
    "    residual_1 = x\n",
    "\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation='relu', padding='same', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation='relu', padding='same', use_bias=False)(x)\n",
    "    x = Dropout(0.10)(x)\n",
    "    x = layers.add([x,residual_1])\n",
    "    \n",
    "    residual_2 = x\n",
    "    \n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = Dropout(0.10)(x)\n",
    "    x = layers.add([x, residual_1, residual_2])\n",
    "    \n",
    "    residual_3 = x\n",
    "    \n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = Conv2D(filters=128, kernel_size=(3, 3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    x = Dropout(0.10)(x)\n",
    "    x = layers.add([x, residual_1, residual_2, residual_3])\n",
    "    \n",
    "    \n",
    "    x = Conv2D(filters=128, kernel_size=(3,3), padding='same', kernel_regularizer=regularization, activation='relu', use_bias=False)(x)\n",
    "    \n",
    "    \"\"\"\n",
    "    x = GlobalAveragePooling2D()(x)\n",
    "    #y = GlobalAveragePooling2D()(residual_1)\n",
    "    #z = GlobalAveragePooling2D()(residual_2)\n",
    "\n",
    "    #x_y_z = layers.add([x,y,z])\n",
    "    \"\"\"\n",
    "    x =Flatten()(x)\n",
    "    \n",
    "    output = Dense(units=3, activation='softmax')(x)\n",
    "    \n",
    "\n",
    "    model = Model(img_input, output)\n",
    "    model.summary()\n",
    "    return model\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "colab_type": "code",
    "id": "imKjZ02DF3VN",
    "outputId": "842628b8-9607-4bc0-fd94-80ebca082444"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"\\n#ALEXNET ARCHITECTURE\\nnp.random.seed(1000)\\n\\n# Constructing CNN structure\\nmodel = Sequential()\\n\\n# 1st convolution layer\\nmodel.add(Conv2D(64, (4, 4), activation='relu', padding = 'valid', input_shape=(256,256,1), bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(Conv2D(64, (4, 4), activation='relu', padding = 'valid', input_shape=(256,256,1), bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(BatchNormalization())\\nmodel.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\\nmodel.add(Dropout(0.25))\\n\\n# 3rd convolution layer\\nmodel.add(Conv2D(64, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(Conv2D(64, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(BatchNormalization())\\nmodel.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\\nmodel.add(Dropout(0.25))\\n          \\n# 5th convolution layer\\nmodel.add(Conv2D(128, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(Conv2D(128, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(BatchNormalization())\\nmodel.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\\nmodel.add(Dropout(0.25))\\n\\n# 7th convolution layer\\nmodel.add(Conv2D(256, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(Conv2D(256, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\\nmodel.add(BatchNormalization())\\nmodel.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\\nmodel.add(Dropout(0.5))\\n\\n\\nmodel.add(Flatten())\\n# Fully connected layers\\nmodel.add(Dense(2048, activation='relu'))\\nmodel.add(Dropout(0.5))\\nmodel.add(Dense(3, activation='softmax'))\\n\""
      ]
     },
     "execution_count": 67,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "'''\n",
    "#ALEXNET ARCHITECTURE\n",
    "np.random.seed(1000)\n",
    "\n",
    "# Constructing CNN structure\n",
    "model = Sequential()\n",
    "\n",
    "# 1st convolution layer\n",
    "model.add(Conv2D(64, (4, 4), activation='relu', padding = 'valid', input_shape=(256,256,1), bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(Conv2D(64, (4, 4), activation='relu', padding = 'valid', input_shape=(256,256,1), bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "# 3rd convolution layer\n",
    "model.add(Conv2D(64, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(Conv2D(64, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "          \n",
    "# 5th convolution layer\n",
    "model.add(Conv2D(128, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(Conv2D(128, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "# 7th convolution layer\n",
    "model.add(Conv2D(256, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(Conv2D(256, (3, 3), activation='relu', padding = 'same', bias_initializer=RandomNormal(stddev=1), kernel_initializer=RandomNormal(stddev=1)))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=(3,3), strides=(2, 2)))\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "\n",
    "model.add(Flatten())\n",
    "# Fully connected layers\n",
    "model.add(Dense(2048, activation='relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(3, activation='softmax'))\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 901
    },
    "colab_type": "code",
    "id": "N0DJthSwr6he",
    "outputId": "3b7746f0-a7cb-4281-9b8b-87a270285068"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_5\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_9 (InputLayer)            [(None, 256, 256, 1) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_55 (Conv2D)              (None, 256, 256, 128 3200        input_9[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_16 (MaxPooling2D) (None, 128, 128, 128 0           conv2d_55[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_23 (Dropout)            (None, 128, 128, 128 0           max_pooling2d_16[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_56 (Conv2D)              (None, 128, 128, 128 147456      dropout_23[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_57 (Conv2D)              (None, 128, 128, 128 147456      conv2d_56[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_24 (Dropout)            (None, 128, 128, 128 0           conv2d_57[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "add_21 (Add)                    (None, 128, 128, 128 0           dropout_24[0][0]                 \n",
      "                                                                 dropout_23[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_58 (Conv2D)              (None, 128, 128, 128 147456      add_21[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_59 (Conv2D)              (None, 128, 128, 128 147456      conv2d_58[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_25 (Dropout)            (None, 128, 128, 128 0           conv2d_59[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "add_22 (Add)                    (None, 128, 128, 128 0           dropout_25[0][0]                 \n",
      "                                                                 dropout_23[0][0]                 \n",
      "                                                                 add_21[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_60 (Conv2D)              (None, 128, 128, 128 147456      add_22[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_61 (Conv2D)              (None, 128, 128, 128 147456      conv2d_60[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_26 (Dropout)            (None, 128, 128, 128 0           conv2d_61[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "add_23 (Add)                    (None, 128, 128, 128 0           dropout_26[0][0]                 \n",
      "                                                                 dropout_23[0][0]                 \n",
      "                                                                 add_21[0][0]                     \n",
      "                                                                 add_22[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_62 (Conv2D)              (None, 128, 128, 128 147456      add_23[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "flatten (Flatten)               (None, 2097152)      0           conv2d_62[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_5 (Dense)                 (None, 3)            6291459     flatten[0][0]                    \n",
      "==================================================================================================\n",
      "Total params: 7,326,851\n",
      "Trainable params: 7,326,851\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input_shape = X.shape[1:]\n",
    "num_classes = 3\n",
    "model = architecture(input_shape, num_classes=3)\n",
    "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "439YlwZFr6hg"
   },
   "outputs": [],
   "source": [
    "filepath='Model.{epoch:02d}-{val_acc:.4f}.hdf5'\n",
    "checkpointer = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode='auto')\n",
    "tensorboard = TensorBoard(log_dir='./logs')\n",
    "learning_rate_reduction = ReduceLROnPlateau(monitor='val_acc', patience=3, verbose=1, factor=0.5, min_lr=0.00001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 391
    },
    "colab_type": "code",
    "id": "XlJb5ledr6hi",
    "outputId": "f4c5c214-5893-41fc-cb10-c6ee9ba22c39"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1499/1500 [============================>.] - ETA: 0s - loss: 3.5076 - acc: 0.4515Epoch 1/10\n",
      " 644/1500 [===========>..................] - ETA: 13s - loss: 1.2596 - acc: 0.4612\n",
      "Epoch 00001: val_loss improved from inf to 1.25655, saving model to Model.01-0.4612.hdf5\n",
      "1500/1500 [==============================] - 1302s 868ms/step - loss: 3.5061 - acc: 0.4515 - val_loss: 1.2566 - val_acc: 0.4612\n",
      "Epoch 2/10\n",
      "1499/1500 [============================>.] - ETA: 0s - loss: 1.5915 - acc: 0.4551Epoch 1/10\n",
      " 644/1500 [===========>..................] - ETA: 12s - loss: 1.2300 - acc: 0.4612\n",
      "Epoch 00002: val_loss improved from 1.25655 to 1.22674, saving model to Model.02-0.4612.hdf5\n",
      "1500/1500 [==============================] - 1298s 865ms/step - loss: 1.5912 - acc: 0.4552 - val_loss: 1.2267 - val_acc: 0.4612\n",
      "Epoch 3/10\n",
      "1499/1500 [============================>.] - ETA: 0s - loss: 1.6247 - acc: 0.4558Epoch 1/10\n",
      " 644/1500 [===========>..................] - ETA: 12s - loss: 1.2136 - acc: 0.4612\n",
      "Epoch 00003: val_loss improved from 1.22674 to 1.21036, saving model to Model.03-0.4612.hdf5\n",
      "1500/1500 [==============================] - 1298s 866ms/step - loss: 1.6244 - acc: 0.4558 - val_loss: 1.2104 - val_acc: 0.4612\n",
      "Epoch 4/10\n",
      "1499/1500 [============================>.] - ETA: 0s - loss: 1.6118 - acc: 0.4545Epoch 1/10\n",
      " 644/1500 [===========>..................] - ETA: 12s - loss: 1.1979 - acc: 0.4612\n",
      "Epoch 00004: val_loss improved from 1.21036 to 1.19465, saving model to Model.04-0.4612.hdf5\n",
      "1500/1500 [==============================] - 1297s 865ms/step - loss: 1.6115 - acc: 0.4545 - val_loss: 1.1946 - val_acc: 0.4612\n",
      "Epoch 5/10\n",
      " 990/1500 [==================>...........] - ETA: 7:18 - loss: 1.3405 - acc: 0.4560"
     ]
    }
   ],
   "source": [
    "# Compiling model\n",
    "model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])\n",
    "\n",
    "\n",
    "\n",
    "# Specifying parameters for Data Augmentation\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",
    "    rotation_range=40,  # randomly rotate images in the range (degrees, 0 to 180)\n",
    "    width_shift_range=0.2,  # randomly shift images horizontally (fraction of total width)\n",
    "    height_shift_range=0.2,  # randomly shift images vertically (fraction of total height)\n",
    "    horizontal_flip=True,  # randomly flip images\n",
    "    vertical_flip=True,\n",
    "    zoom_range = 0.05)  # zoom images in range [1 - zoom_range, 1+ zoom_range]\n",
    "\n",
    "datagen.fit(X_train)\n",
    "\n",
    "\n",
    "# Saving model each time it achieves lower loss on the validation set\n",
    "filepath='Model.{epoch:02d}-{val_acc:.4f}.hdf5'\n",
    "checkpointer = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode='auto')\n",
    "tensorboard = TensorBoard(log_dir='./logs')\n",
    "\n",
    "\n",
    "history = model.fit_generator(datagen.flow(X_train, y_train,\n",
    "                    batch_size=16),\n",
    "                    epochs=10,\n",
    "                    validation_data=(X_val, y_val),\n",
    "                    steps_per_epoch=X_train.shape[0],\n",
    "                    callbacks=[checkpointer,tensorboard]\n",
    "                             )\n",
    "\n",
    "\n",
    "pd.DataFrame(history.history).to_csv(\"history.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "QyvL7sLir6hk"
   },
   "outputs": [],
   "source": [
    "model2 = load_model('Model.05-0.4791.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "lylh12iCONI1",
    "outputId": "44536e24-3a8a-491d-bc46-6e7a21f1f22b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "920/920 [==============================] - 12s 13ms/sample - loss: 1.0533 - acc: 0.4837\n",
      "Loss: 1.0532933297364608\n",
      "Accuracy: 0.48369566\n"
     ]
    }
   ],
   "source": [
    "scores = model.evaluate(np.array(X_test), np.array(y_test), batch_size=16)\n",
    "print(\"Loss: \" + str(scores[0]))\n",
    "print(\"Accuracy: \" + str(scores[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "XoHIaBmyONyL"
   },
   "outputs": [],
   "source": [
    "history = pd.read_csv('history.csv', usecols = ['acc','loss','val_acc','val_loss'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 638
    },
    "colab_type": "code",
    "id": "G3O3-cp-OPBD",
    "outputId": "41a08bda-2bf9-468b-e83f-203dfc34e368"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABLsAAAJtCAYAAADesXLzAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_accuracy(data, size = (20,10)):\n",
    "    plt.figure(figsize=size) \n",
    "    plt.plot(data['acc'])\n",
    "    plt.plot(data['val_acc'])\n",
    "    plt.title('Model Accuracy', fontsize = 18)\n",
    "    plt.ylabel('Accuracy', fontsize = 18)\n",
    "    plt.xlabel('Epoch', fontsize = 18)\n",
    "    plt.legend(['Train', 'Test'], loc='upper left', fontsize = 16)\n",
    "    plt.tick_params(axis='both', labelsize=14)\n",
    "    plt.show()\n",
    "    \n",
    "plot_accuracy(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 638
    },
    "colab_type": "code",
    "id": "q9TcXfOzOaGa",
    "outputId": "c4c87661-3b1c-4928-9a46-c4afd4a27365"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABKEAAAJtCAYAAADuPWPSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSWru/wEc861JzEogyAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_loss(data, size = (20,10)):\n",
    "    plt.figure(figsize=size) \n",
    "    plt.plot(data['loss'])\n",
    "    plt.plot(data['val_loss'])\n",
    "    plt.title('Model Loss', fontsize = 18)\n",
    "    plt.ylabel('Loss', fontsize = 18)\n",
    "    plt.xlabel('Epoch', fontsize = 18)\n",
    "    plt.ylim(0.9,2)\n",
    "    plt.legend(['Train', 'Test'], loc='upper left', fontsize = 16)\n",
    "    plt.tick_params(axis='both', labelsize=14)\n",
    "    plt.show()\n",
    "    \n",
    "plot_loss(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "eSiXeUF5OabS"
   },
   "outputs": [],
   "source": [
    "def create_confmat(true_labels, predicted_labels, columns, colour = 'Oranges', size = (20,14)):\n",
    "    \n",
    "    cm = confusion_matrix(true_labels, predicted_labels) \n",
    "    cm_df = pd.DataFrame(cm,\n",
    "    index = [col for col in columns], \n",
    "    columns = [col for col in columns])\n",
    "    plt.figure(figsize=(18,16))\n",
    "    sns.heatmap(cm_df, annot = True, cmap = colour, fmt='g', linewidths=.2)\n",
    "    plt.title('Confusion Matrix', fontsize = 20)\n",
    "    plt.ylabel('True label', fontsize = 18)\n",
    "    plt.xlabel('Predicted label', fontsize = 18)\n",
    "    plt.tick_params(axis='both', labelsize=14)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 163
    },
    "colab_type": "code",
    "id": "r_xb5NluOfrD",
    "outputId": "8ad935f3-4549-4b29-ba0b-edd1f395224a"
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "ignored",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-45-53a6e0884fc6>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mcreate_confmat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mresults_df\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'Original_label'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresults_df\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'Predicted_label'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m'GLIOMA'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'MENINGIOMA'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'PITUITARY'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcolour\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m'Greens'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'results_df' is not defined"
     ]
    }
   ],
   "source": [
    "# Displaying confusion matrix\n",
    "create_confmat(results_df['Original_label'], results_df['Predicted_label'], ['GLIOMA','MENINGIOMA','PITUITARY'], colour = 'Greens')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "zdeRs83jPmKN"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "train_model.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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}