breast_cancer_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\chandrachurh\\appdata\\local\\programs\\python\\python35\\lib\\importlib\\_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88\n",
      "  return f(*args, **kwds)\n",
      "c:\\users\\chandrachurh\\appdata\\local\\programs\\python\\python35\\lib\\importlib\\_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88\n",
      "  return f(*args, **kwds)\n",
      "c:\\users\\chandrachurh\\appdata\\local\\programs\\python\\python35\\lib\\importlib\\_bootstrap.py:222: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88\n",
      "  return f(*args, **kwds)\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from pathlib import Path\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import json\n",
    "import glob\n",
    "import os\n",
    "import os\n",
    "import cv2\n",
    "import glob\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import random\n",
    "import pickle\n",
    "from tqdm import tqdm\n",
    "from pathlib import Path\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.image as mpimg\n",
    "from mlxtend.plotting import plot_confusion_matrix\n",
    "from scipy import ndimage\n",
    "import skimage\n",
    "from skimage.io import imread\n",
    "from skimage.transform import resize\n",
    "from sklearn.utils import shuffle\n",
    "from sklearn.metrics import confusion_matrix, precision_score, recall_score\n",
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.model_selection import StratifiedShuffleSplit as s_split\n",
    "import tensorflow as tf\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense , Activation\n",
    "from keras.layers import Dropout\n",
    "from keras.layers import Flatten\n",
    "from keras.constraints import maxnorm\n",
    "from keras.optimizers import SGD , RMSprop, Adam\n",
    "from keras.layers import Conv2D , SeparableConv2D, BatchNormalization, UpSampling2D, MaxPool2D\n",
    "from keras.layers import MaxPooling2D, Average, Input, Concatenate, LeakyReLU, Add, ELU, PReLU, ReLU\n",
    "from keras.callbacks import EarlyStopping\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.callbacks import ReduceLROnPlateau , ModelCheckpoint\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from keras import backend as BE\n",
    "from vis.visualization import visualize_cam, visualize_cam_with_losses, visualize_activation, visualize_saliency\n",
    "from vis.utils import utils\n",
    "from keras.applications.inception_v3 import InceptionV3\n",
    "from keras.applications.inception_resnet_v2 import InceptionResNetV2\n",
    "from keras.applications.xception import Xception\n",
    "from keras.applications.resnet50 import ResNet50\n",
    "from keras.layers import GlobalAveragePooling2D\n",
    "from keras.models import Model\n",
    "from keras.utils import plot_model\n",
    "from numpy.random import seed\n",
    "from tensorflow import set_random_seed"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font size=5>The data processing</font>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "def set_data(path):\n",
    "    for next_dir in os.listdir(path):\n",
    "        temp = path + next_dir\n",
    "        for dirc in os.listdir(temp):\n",
    "            temp1 = path +next_dir+'/'+dirc\n",
    "            for file in tqdm(os.listdir(temp1)):\n",
    "                img = cv2.imread(temp1+\"/\"+file)\n",
    "                img = cv2.resize(img, (50,50))\n",
    "                if dirc ==\"0\":\n",
    "                    cv2.imwrite(\"data/NORMAL/\"+file,img)\n",
    "                else:\n",
    "                    cv2.imwrite(\"data/IDC/\"+file,img)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "scrolled": true
   },
   "source": [
    "set_data(\"kaggle_data/\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font size=5>The Model Development</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_image_data(path):\n",
    "    X=[]\n",
    "    y=[]\n",
    "    for nextDir in os.listdir(path):\n",
    "        if not nextDir.startswith('.'):\n",
    "            if nextDir in ['NORMAL']:\n",
    "                label = 0\n",
    "            elif nextDir in ['IDC']:\n",
    "                label = 1\n",
    "            else:\n",
    "                label = 2\n",
    "\n",
    "            temp = path + nextDir\n",
    "\n",
    "            for file in tqdm(os.listdir(temp)):\n",
    "                img = cv2.imread(temp + '/' + file, 1)\n",
    "                img = cv2.resize(img,(100,100))\n",
    "                img1 = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2LAB))[0]\n",
    "                img2 = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))[0]\n",
    "                img3 = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))[1]\n",
    "                img4 = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2LAB))[1]\n",
    "                if img is not None:\n",
    "                    clahe = cv2.createCLAHE(clipLimit=5, tileGridSize=(16,16))\n",
    "                    planes = cv2.split(img)\n",
    "                    for i in range(0,3):\n",
    "                        planes[i] =clahe.apply(planes[i])\n",
    "                    img = cv2.merge(planes)\n",
    "                    #img = clahe.apply(img)\n",
    "                    img = cv2.GaussianBlur(img,(5,5),0)\n",
    "                    #img = skimage.transform.resize(img, (299, 299, 3))\n",
    "                    img = cv2.merge([img,img1,img2,img3,img4])\n",
    "                    img = np.asarray(img)\n",
    "                    X.append(img)\n",
    "                    y.append(label)\n",
    "\n",
    "    X = np.asarray(X)\n",
    "    y = np.asarray(y)\n",
    "    return X,y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████████████| 3000/3000 [00:08<00:00, 365.75it/s]\n",
      "100%|█████████████████████████████████████████████████████████████████████████████| 4500/4500 [00:11<00:00, 408.87it/s]\n"
     ]
    }
   ],
   "source": [
    "X, y = get_image_data(\"fur_fur_fur_reduced_data/\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def shuffle_data(X_shu, y_shu):\n",
    "    #X_shu=np.asarray(X_shu)\n",
    "    #y_shu=np.asarray(y_shu)\n",
    "    split = s_split(n_splits= 1, test_size = 0.15, random_state=18)\n",
    "    X_train = []\n",
    "    y_train = []\n",
    "    X_test = []\n",
    "    y_test = []\n",
    "    for train_id, test_id in split.split(X_shu, y_shu):\n",
    "        X_train.append(X_shu[train_id])\n",
    "        y_train.append(y_shu[train_id])\n",
    "        X_test.append(X_shu[test_id])\n",
    "        y_test.append(y_shu[test_id])\n",
    "    #del X_shu\n",
    "    #del y_shu\n",
    "    return X_train, y_train, X_test, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, y_train, X_test, y_test = shuffle_data(X,y)\n",
    "X_train = np.asarray(X_train, dtype = \"float32\")[0]\n",
    "X_test = np.asarray(X_test, dtype = \"float32\")[0]\n",
    "y_train = np.asarray(y_train)[0]\n",
    "y_test = np.asarray(y_test)[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train /= 255\n",
    "X_test /= 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test = to_categorical(y_test,2)\n",
    "y_train = to_categorical(y_train,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1125, 100, 100, 7)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font size=5>The Model</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "BE.clear_session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def residual_block(mod_, f_in, f_out, strides_ = (1,1), use_shortcut_ = False):   \n",
    "    shortcut_ = mod_\n",
    "    \n",
    "    k_ = (1,1)\n",
    "    \n",
    "    mod_ = Conv2D(f_in, kernel_size=k_, strides=(1,1), padding = \"same\")(mod_)\n",
    "    mod_ = BatchNormalization()(mod_)\n",
    "    mod_ = ELU()(mod_)\n",
    "    \n",
    "    mod_ = Conv2D(f_in, kernel_size=k_, strides=strides_, padding = \"same\")(mod_)\n",
    "    mod_ = BatchNormalization()(mod_)\n",
    "    mod_ = ELU()(mod_)\n",
    "    \n",
    "    mod_ = Conv2D(f_out, kernel_size=k_, strides=(1,1), padding = \"same\")(mod_)\n",
    "    mod_ = BatchNormalization()(mod_)\n",
    "    mod_ = ELU()(mod_)\n",
    "    \n",
    "    if use_shortcut_ == True or strides_ != (1,1):\n",
    "        shortcut_ = Conv2D(f_out, kernel_size=k_, strides=strides_, padding = \"same\")(shortcut_)\n",
    "        shortcut_ = BatchNormalization()(shortcut_)\n",
    "        \n",
    "    mod_ = Add()([shortcut_, mod_])\n",
    "    mod_ = ReLU()(mod_)\n",
    "    \n",
    "    return mod_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model_build(in_):\n",
    "    \n",
    "    k_=(1,1)\n",
    "    \n",
    "    mod_ = Conv2D(16, kernel_size=k_, strides = (1,1), padding = \"same\")(in_)\n",
    "    \n",
    "    mod_ = BatchNormalization()(mod_)\n",
    "    \n",
    "    mod_ = ReLU()(mod_)\n",
    "    \n",
    "    mod_ = MaxPooling2D()(mod_)\n",
    "    \n",
    "    mod_ = residual_block(mod_, 16, 32, use_shortcut_=True)\n",
    "    \n",
    "    mod_ = MaxPooling2D()(mod_)\n",
    "    \n",
    "    mod_ = residual_block(mod_, 32, 64, use_shortcut_=True)\n",
    "    \n",
    "    mod_ = MaxPooling2D()(mod_)\n",
    "    \n",
    "    mod_ = residual_block(mod_, 64, 96, use_shortcut_=True)\n",
    "    \n",
    "    mod_ = MaxPooling2D()(mod_)\n",
    "    \n",
    "    mod_ = residual_block(mod_, 96, 128, use_shortcut_=True)\n",
    "    \n",
    "    mod_ = MaxPooling2D()(mod_)\n",
    "    \n",
    "    mod_ = GlobalAveragePooling2D()(mod_)\n",
    "    \n",
    "    mod_ = Dense(512, activation = \"relu\")(mod_)\n",
    "    \n",
    "    mod_ = Dropout(0.5)(mod_)\n",
    "    \n",
    "    mod_ = Dense(2, activation=\"softmax\")(mod_)\n",
    "    \n",
    "    return mod_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "in_ = Input((100,100,7))\n",
    "model = model_build(in_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\chandrachurh\\appdata\\local\\programs\\python\\python35\\lib\\site-packages\\ipykernel_launcher.py:1: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=Tensor(\"in..., outputs=Tensor(\"de...)`\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    }
   ],
   "source": [
    "model_f = Model(input = in_, output = model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 100, 100, 7)  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 100, 100, 16) 128         input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 100, 100, 16) 64          conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_1 (ReLU)                  (None, 100, 100, 16) 0           batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 50, 50, 16)   0           re_lu_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 50, 50, 16)   272         max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 50, 50, 16)   64          conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "elu_1 (ELU)                     (None, 50, 50, 16)   0           batch_normalization_2[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_3 (Conv2D)               (None, 50, 50, 16)   272         elu_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 50, 50, 16)   64          conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "elu_2 (ELU)                     (None, 50, 50, 16)   0           batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_4 (Conv2D)               (None, 50, 50, 32)   544         elu_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_5 (Conv2D)               (None, 50, 50, 32)   544         max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_4 (BatchNor (None, 50, 50, 32)   128         conv2d_4[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_5 (BatchNor (None, 50, 50, 32)   128         conv2d_5[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "elu_3 (ELU)                     (None, 50, 50, 32)   0           batch_normalization_4[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "add_1 (Add)                     (None, 50, 50, 32)   0           batch_normalization_5[0][0]      \n",
      "                                                                 elu_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_2 (ReLU)                  (None, 50, 50, 32)   0           add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2D)  (None, 25, 25, 32)   0           re_lu_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_6 (Conv2D)               (None, 25, 25, 32)   1056        max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_6 (BatchNor (None, 25, 25, 32)   128         conv2d_6[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "elu_4 (ELU)                     (None, 25, 25, 32)   0           batch_normalization_6[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_7 (Conv2D)               (None, 25, 25, 32)   1056        elu_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_7 (BatchNor (None, 25, 25, 32)   128         conv2d_7[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "elu_5 (ELU)                     (None, 25, 25, 32)   0           batch_normalization_7[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_8 (Conv2D)               (None, 25, 25, 64)   2112        elu_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 25, 25, 64)   2112        max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_8 (BatchNor (None, 25, 25, 64)   256         conv2d_8[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_9 (BatchNor (None, 25, 25, 64)   256         conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "elu_6 (ELU)                     (None, 25, 25, 64)   0           batch_normalization_8[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 25, 25, 64)   0           batch_normalization_9[0][0]      \n",
      "                                                                 elu_6[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_3 (ReLU)                  (None, 25, 25, 64)   0           add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2D)  (None, 12, 12, 64)   0           re_lu_3[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_10 (Conv2D)              (None, 12, 12, 64)   4160        max_pooling2d_3[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_10 (BatchNo (None, 12, 12, 64)   256         conv2d_10[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "elu_7 (ELU)                     (None, 12, 12, 64)   0           batch_normalization_10[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_11 (Conv2D)              (None, 12, 12, 64)   4160        elu_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_11 (BatchNo (None, 12, 12, 64)   256         conv2d_11[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "elu_8 (ELU)                     (None, 12, 12, 64)   0           batch_normalization_11[0][0]     \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "conv2d_12 (Conv2D)              (None, 12, 12, 96)   6240        elu_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_13 (Conv2D)              (None, 12, 12, 96)   6240        max_pooling2d_3[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_12 (BatchNo (None, 12, 12, 96)   384         conv2d_12[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_13 (BatchNo (None, 12, 12, 96)   384         conv2d_13[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "elu_9 (ELU)                     (None, 12, 12, 96)   0           batch_normalization_12[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "add_3 (Add)                     (None, 12, 12, 96)   0           batch_normalization_13[0][0]     \n",
      "                                                                 elu_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_4 (ReLU)                  (None, 12, 12, 96)   0           add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2D)  (None, 6, 6, 96)     0           re_lu_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_14 (Conv2D)              (None, 6, 6, 96)     9312        max_pooling2d_4[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_14 (BatchNo (None, 6, 6, 96)     384         conv2d_14[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "elu_10 (ELU)                    (None, 6, 6, 96)     0           batch_normalization_14[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_15 (Conv2D)              (None, 6, 6, 96)     9312        elu_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_15 (BatchNo (None, 6, 6, 96)     384         conv2d_15[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "elu_11 (ELU)                    (None, 6, 6, 96)     0           batch_normalization_15[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_16 (Conv2D)              (None, 6, 6, 128)    12416       elu_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_17 (Conv2D)              (None, 6, 6, 128)    12416       max_pooling2d_4[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_16 (BatchNo (None, 6, 6, 128)    512         conv2d_16[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_17 (BatchNo (None, 6, 6, 128)    512         conv2d_17[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "elu_12 (ELU)                    (None, 6, 6, 128)    0           batch_normalization_16[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "add_4 (Add)                     (None, 6, 6, 128)    0           batch_normalization_17[0][0]     \n",
      "                                                                 elu_12[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "re_lu_5 (ReLU)                  (None, 6, 6, 128)    0           add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2D)  (None, 3, 3, 128)    0           re_lu_5[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_1 (Glo (None, 128)          0           max_pooling2d_5[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 512)          66048       global_average_pooling2d_1[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dropout_1 (Dropout)             (None, 512)          0           dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 2)            1026        dropout_1[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 143,714\n",
      "Trainable params: 141,570\n",
      "Non-trainable params: 2,144\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model_f.compile(optimizer = RMSprop(), loss = \"binary_crossentropy\", metrics=[\"accuracy\"])\n",
    "model_f.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font size=5>Specifying parameters</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "nb_epochs = 50\n",
    "nb_batch = 20\n",
    "\n",
    "earlyStopping=EarlyStopping(monitor='val_loss', patience=5, verbose=0, mode='auto')\n",
    "\n",
    "lr_reduce = ReduceLROnPlateau(monitor='val_acc', factor=0.1, epsilon=0.0001, patience=2, verbose=1)\n",
    "\n",
    "save_path=\"weights/res5_7x7_BGR_l_h_s_a_clahe_20_blur_new_fur_fur_fur_reddata_ELU_acti_convo_test_lastelu.h5\"\n",
    "\n",
    "checkpoint = ModelCheckpoint(save_path, monitor='val_acc', verbose=1, save_best_only=True, mode='max')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "history=model_f.fit(X_train, y_train, epochs = nb_epochs, batch_size = nb_batch, callbacks=[earlyStopping, lr_reduce, checkpoint], validation_data=(X_test, y_test), verbose = 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('history/res4_1x1_BGR_l_h_s_a_clahe_20_blur_new_fur_fur_fur_reddata_ELU_acti_convo_test_lastelu', 'wb') as file_pi:\n",
    "    pickle.dump(history.history, file_pi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_f.load_weights(\"weights/res4_1x1_BGR_l_h_s_a_clahe_20_blur_new_fur_fur_fur_reddata_ELU_acti_convo_test_lastelu.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.022613801583406208, 0.9937777777777778]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_f.evaluate(X_test,y_test,verbose=2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font  size=5>Visualization</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x261d0c1e4a8>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "idx=13\n",
    "plt.imshow(cv2.merge(cv2.split(X_test[idx])[0:3]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = np.expand_dims(X_test[idx], axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 0s 266ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_f.predict(x, verbose=1)[0].argmax(axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1., 0.], dtype=float32)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test[idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x261d0ce4978>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(visualize_cam(model=model_f, layer_idx=utils.find_layer_idx(model_f, 're_lu_4'), filter_indices=None, seed_input=np.asarray(X_test[idx]), penultimate_layer_idx=utils.find_layer_idx(model_f, 'add_3'), \\\n",
    "    backprop_modifier=None, grad_modifier=None))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font size =  5>The performance metrics</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9889624724061811\n",
      "0.9955555555555555\n"
     ]
    }
   ],
   "source": [
    "y_test_labels = np.argmax(y_test, axis = 1)\n",
    "y_test_pred_incep = np.argmax(model_f.predict(X_test),axis = 1)\n",
    "print(precision_score(y_test_labels, y_test_pred_incep))\n",
    "print(recall_score(y_test_labels, y_test_pred_incep))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "0.9996411522633745"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_probs = model_f.predict([X_test])\n",
    "fpr, tpr, thresholds = roc_curve(y_test_labels, y_pred_probs[:, 1])\n",
    "def plot_roc_curve(fpr, tpr, label = None):\n",
    "    plt.plot(fpr, tpr, linewidth=2, label = label)\n",
    "    plt.plot([0, 1], [0, 1], \"k--\")\n",
    "    plt.axis([0, 1, 0, 1])\n",
    "    plt.xlabel(\"False Positive Rate\")\n",
    "    plt.xlabel(\"True Positive Rate\")\n",
    "\n",
    "plot_roc_curve(fpr, tpr)\n",
    "plt.show()\n",
    "roc_auc_score(y_test_labels,y_pred_probs[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['loss', 'lr', 'val_loss', 'acc', 'val_acc'])\n"
     ]
    }
   ],
   "source": [
    "filename = open('history/res3_3x3_BGR_l_h_s_a_clahe_24_blur_new_fur_fur_reddata_acti',\"rb\")\n",
    "history = pickle.load(filename)\n",
    "print(history.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history['acc'])\n",
    "plt.plot(history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history['loss'])\n",
    "plt.plot(history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}