Pneumonia_Classification_Model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ukb7nZjIMQcy"
   },
   "source": [
    "### Author: [Pratik Sharma](https://github.com/sharmapratik88/)\n",
    "## Import Packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 122
    },
    "colab_type": "code",
    "id": "al8SUKiPIOkW",
    "outputId": "a98e6ffb-cf96-429a-ee46-b32633290cc8",
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>.container { width:100% !important; }</style>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from IPython.core.display import display, HTML\n",
    "display(HTML('<style>.container { width:100% !important; }</style>'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "5rGulN57IXb0",
    "outputId": "6302d6db-1ef9-4604-f04b-ac942dcb0a08"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Device mapping:\n",
      "/job:localhost/replica:0/task:0/device:XLA_CPU:0 -> device: XLA_CPU device\n",
      "/job:localhost/replica:0/task:0/device:XLA_GPU:0 -> device: XLA_GPU device\n",
      "/job:localhost/replica:0/task:0/device:GPU:0 -> device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5\n",
      "\n",
      "Num GPUs Available:  1\n",
      "Mixed precision enabled\n"
     ]
    }
   ],
   "source": [
    "# IMPORT PACKAGES\n",
    "import warnings; warnings.filterwarnings('ignore')\n",
    "import pandas as pd, numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import os, random, csv\n",
    "import seaborn as sns\n",
    "from glob import glob\n",
    "from tqdm import tqdm\n",
    "%matplotlib inline\n",
    "\n",
    "# MODEL METRICS\n",
    "from mlxtend.plotting import plot_confusion_matrix\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# SET VARIABLES\n",
    "PATH = '/home/pratik_sharma_ds/Capstone_Pneumonia_Detection/'\n",
    "os.chdir(PATH)\n",
    "\n",
    "DATA_DIR = os.path.join(PATH + 'data/')\n",
    "TRAIN_IMAGES = os.path.join(DATA_DIR + 'stage_2_train_images/')\n",
    "TEST_IMAGES = os.path.join(DATA_DIR + 'stage_2_test_images/')\n",
    "OUTPUT_DIR = os.path.join(PATH + 'output/')\n",
    "MODEL_WEIGHTS = os.path.join(PATH + 'model_weights/')\n",
    "TRAIN_IMAGES_DIR = os.path.join(DATA_DIR + 'stage_2_train_images_png/')\n",
    "if not os.path.exists(TRAIN_IMAGES_DIR): os.makedirs(TRAIN_IMAGES_DIR)\n",
    "if not os.path.exists(MODEL_WEIGHTS): os.makedirs(MODEL_WEIGHTS)\n",
    "\n",
    "# CUSTOM MODULE\n",
    "from module.classify import *\n",
    "\n",
    "# SET RANDOM SEED\n",
    "random_state = 2020\n",
    "random.seed(random_state)\n",
    "np.random.seed(random_state)\n",
    "\n",
    "# USE GPU FOR MODEL TRAINING\n",
    "config = tf.compat.v1.ConfigProto(log_device_placement = True, allow_soft_placement = True)\n",
    "config.gpu_options.allow_growth = True\n",
    "session = tf.compat.v1.Session(config = config)\n",
    "list_gpu = tf.config.experimental.list_physical_devices(device_type = 'GPU')\n",
    "if len(list_gpu)==1: strategy = tf.distribute.OneDeviceStrategy(device='/gpu:0')\n",
    "else: strategy = tf.distribute.MirroredStrategy()\n",
    "print('Num GPUs Available: ', len(list_gpu))\n",
    "tf.config.optimizer.set_experimental_options({'auto_mixed_precision': True})\n",
    "print('Mixed precision enabled')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "X5ZtjK4MMaYq"
   },
   "source": [
    "## Read Data & Prep for Model\n",
    "* Read a pickle file that was saved as part of [EDA & Data Prep](https://github.com/sharmapratik88/Capstone_Pneumonia_Detection/blob/master/Pneumonia_Detection_EDA_%26_Data_Prep.ipynb)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 306
    },
    "colab_type": "code",
    "id": "-baCLrK_u7Zg",
    "outputId": "26732056-9da9-4645-d8c4-9ba8ddf2fc1e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(30227, 12)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>patientId</th>\n",
       "      <th>x</th>\n",
       "      <th>y</th>\n",
       "      <th>width</th>\n",
       "      <th>height</th>\n",
       "      <th>Target</th>\n",
       "      <th>number_of_boxes</th>\n",
       "      <th>class</th>\n",
       "      <th>path</th>\n",
       "      <th>PatientSex</th>\n",
       "      <th>PatientAge</th>\n",
       "      <th>ViewPosition</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0004cfab-14fd-4e49-80ba-63a80b6bddd6</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>No Lung Opacity / Not Normal</td>\n",
       "      <td>/home/pratik_sharma_ds/Capstone_Pneumonia_Dete...</td>\n",
       "      <td>F</td>\n",
       "      <td>51.0</td>\n",
       "      <td>PA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>00313ee0-9eaa-42f4-b0ab-c148ed3241cd</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>No Lung Opacity / Not Normal</td>\n",
       "      <td>/home/pratik_sharma_ds/Capstone_Pneumonia_Dete...</td>\n",
       "      <td>F</td>\n",
       "      <td>48.0</td>\n",
       "      <td>PA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>00322d4d-1c29-4943-afc9-b6754be640eb</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>No Lung Opacity / Not Normal</td>\n",
       "      <td>/home/pratik_sharma_ds/Capstone_Pneumonia_Dete...</td>\n",
       "      <td>M</td>\n",
       "      <td>19.0</td>\n",
       "      <td>AP</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>003d8fa0-6bf1-40ed-b54c-ac657f8495c5</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>Normal</td>\n",
       "      <td>/home/pratik_sharma_ds/Capstone_Pneumonia_Dete...</td>\n",
       "      <td>M</td>\n",
       "      <td>28.0</td>\n",
       "      <td>PA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>00436515-870c-4b36-a041-de91049b9ab4</td>\n",
       "      <td>264.0</td>\n",
       "      <td>152.0</td>\n",
       "      <td>213.0</td>\n",
       "      <td>379.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>Lung Opacity</td>\n",
       "      <td>/home/pratik_sharma_ds/Capstone_Pneumonia_Dete...</td>\n",
       "      <td>F</td>\n",
       "      <td>32.0</td>\n",
       "      <td>AP</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                              patientId      x      y  width  height  Target  \\\n",
       "0  0004cfab-14fd-4e49-80ba-63a80b6bddd6    NaN    NaN    NaN     NaN       0   \n",
       "1  00313ee0-9eaa-42f4-b0ab-c148ed3241cd    NaN    NaN    NaN     NaN       0   \n",
       "2  00322d4d-1c29-4943-afc9-b6754be640eb    NaN    NaN    NaN     NaN       0   \n",
       "3  003d8fa0-6bf1-40ed-b54c-ac657f8495c5    NaN    NaN    NaN     NaN       0   \n",
       "4  00436515-870c-4b36-a041-de91049b9ab4  264.0  152.0  213.0   379.0       1   \n",
       "\n",
       "   number_of_boxes                         class  \\\n",
       "0                1  No Lung Opacity / Not Normal   \n",
       "1                1  No Lung Opacity / Not Normal   \n",
       "2                1  No Lung Opacity / Not Normal   \n",
       "3                1                        Normal   \n",
       "4                2                  Lung Opacity   \n",
       "\n",
       "                                                path PatientSex  PatientAge  \\\n",
       "0  /home/pratik_sharma_ds/Capstone_Pneumonia_Dete...          F        51.0   \n",
       "1  /home/pratik_sharma_ds/Capstone_Pneumonia_Dete...          F        48.0   \n",
       "2  /home/pratik_sharma_ds/Capstone_Pneumonia_Dete...          M        19.0   \n",
       "3  /home/pratik_sharma_ds/Capstone_Pneumonia_Dete...          M        28.0   \n",
       "4  /home/pratik_sharma_ds/Capstone_Pneumonia_Dete...          F        32.0   \n",
       "\n",
       "  ViewPosition  \n",
       "0           PA  \n",
       "1           PA  \n",
       "2           AP  \n",
       "3           PA  \n",
       "4           AP  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Read `train_class_features.pkl` file, saved as EDA & Data Prep Module\n",
    "train_class = pd.read_pickle(OUTPUT_DIR + 'train_class_features.pkl')\n",
    "display(train_class.shape, train_class.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 306
    },
    "colab_type": "code",
    "id": "jy7cjyGivFbU",
    "outputId": "03777498-f830-4585-e57d-d697089f0787"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prepare a dataframe with paths, classes and targets\n",
      "--------------------------------------------------------------------------------\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(26684, 4)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "patientId    26684\n",
       "path         26684\n",
       "class            3\n",
       "Target           2\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Distribution of target and classes\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    20672\n",
       "1     6012\n",
       "Name: Target, dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "No Lung Opacity / Not Normal    11821\n",
       "Normal                           8851\n",
       "Lung Opacity                     6012\n",
       "Name: class, dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# A dataframe with paths, classes and targets\n",
    "print('Prepare a dataframe with paths, classes and targets'); print('--'*40)\n",
    "path_class_target = train_class[['patientId', 'path', 'class', 'Target']].copy(deep = True)\n",
    "path_class_target['path'] = (path_class_target['path']\n",
    "                             .str.replace('stage_2_train_images', 'stage_2_train_images_png')\n",
    "                             .str.replace('.dcm', '.png'))\n",
    "path_class_target.drop_duplicates(inplace = True)\n",
    "display(path_class_target.shape, path_class_target.nunique())\n",
    "print('\\nDistribution of target and classes')\n",
    "display(path_class_target['Target'].value_counts())\n",
    "print()\n",
    "display(path_class_target['class'].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Convert dicom to png images\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "# Run only if you haven't converted to png images yet\n",
    "print('Convert dicom to png images'); print('--'*40)\n",
    "\n",
    "# image_lists = [f for f in os.listdir(TRAIN_IMAGES)]\n",
    "# for f in tqdm_notebook(image_lists): \n",
    "#     ds = dcm.read_file(TRAIN_IMAGES + f)\n",
    "#     img = ds.pixel_array\n",
    "#     cv2.imwrite(TRAIN_IMAGES_DIR + f.replace('.dcm','.png'), img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Split the data in train, valid and test sets\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "# Split the data in train, valid and test sets\n",
    "print('Split the data in train, valid and test sets'); print('--'*40)\n",
    "\n",
    "image_list = list(path_class_target['path'])\n",
    "random.shuffle(image_list)\n",
    "val_size = round(len(image_list)/10)\n",
    "test_size = round(len(image_list)/10)\n",
    "train_size = len(image_list)-test_size-val_size\n",
    "\n",
    "X_train = image_list[:train_size]\n",
    "X_valid = image_list[train_size:(train_size + val_size)]\n",
    "X_test = image_list[(train_size + val_size):]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 119
    },
    "colab_type": "code",
    "id": "q_6QOfAkzPU7",
    "outputId": "1a707eed-44e8-4cfa-d738-95fb2c7c8701"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Create Training, Validation and Test Dataframe with Path and Target\n",
      "--------------------------------------------------------------------------------\n",
      "Shape of the dataframes:\n",
      "TRAIN:(21348, 3)\n",
      "VALID:(2668, 3)\n",
      "TEST:(2668, 3)\n"
     ]
    }
   ],
   "source": [
    "print('Create Training, Validation and Test Dataframe with Path and Target'); print('--'*40)\n",
    "\n",
    "df_train = (path_class_target.merge(pd.Series(X_train, name = 'path'), \n",
    "                                    on = 'path', \n",
    "                                    how = 'right')\n",
    "          .drop(['class'], axis = 1))\n",
    "\n",
    "df_valid = (path_class_target.merge(pd.Series(X_valid, name = 'path'), \n",
    "                                    on = 'path', \n",
    "                                    how = 'right')\n",
    "          .drop(['class'], axis = 1))\n",
    "\n",
    "df_test = (path_class_target.merge(pd.Series(X_test, name = 'path'), \n",
    "                                    on = 'path', \n",
    "                                    how = 'right')\n",
    "          .drop(['class'], axis = 1))\n",
    "\n",
    "print('Shape of the dataframes:\\nTRAIN:{}\\nVALID:{}\\nTEST:{}'.format(df_train.shape, df_valid.shape, df_test.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 289
    },
    "colab_type": "code",
    "id": "FcqHp-blx4IJ",
    "outputId": "20e863c4-403c-49cf-a22d-4114004da0c3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training, Validation and Test set is ~equally distributed on target\n",
      "--------------------------------------------------------------------------------\n",
      "Distribution of target in the training set:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    0.78\n",
       "1    0.22\n",
       "Name: Target, dtype: float64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Distribution of target in the validation set:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    0.78\n",
       "1    0.22\n",
       "Name: Target, dtype: float64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Distribution of target in the test set:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    0.77\n",
       "1    0.23\n",
       "Name: Target, dtype: float64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Training, Validation and Test set is ~equally distributed on target'); print('--'*40)\n",
    "print('Distribution of target in the training set:'); \n",
    "display(pd.Series(df_train['Target']).value_counts(normalize = True).round(2))\n",
    "print('\\nDistribution of target in the validation set:'); \n",
    "display(pd.Series(df_valid['Target']).value_counts(normalize = True).round(2))\n",
    "print('\\nDistribution of target in the test set:'); \n",
    "display(pd.Series(df_test['Target']).value_counts(normalize = True).round(2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "Wu3Z72Djc8pr",
    "outputId": "b6ed505f-a43c-432e-f8bb-b30f3854cfa7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Save the train, valid and test dataframes for future use\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "print('Save the train, valid and test dataframes for future use');print('--'*40)\n",
    "df_train.to_pickle(OUTPUT_DIR + 'train_data.pkl')\n",
    "df_valid.to_pickle(OUTPUT_DIR + 'valid_data.pkl')\n",
    "df_test.to_pickle(OUTPUT_DIR + 'test_data.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "9lVndB3tNpdj"
   },
   "source": [
    "## Model - DenseNet121"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MVGCIGX6Jqqp"
   },
   "outputs": [],
   "source": [
    "# Model Parameters\n",
    "BATCH_SIZE = 32\n",
    "IMAGE_SIZE = 224\n",
    "EPOCH = 10\n",
    "LEARNING_RATE = 1e-4\n",
    "MONITOR = 'val_loss'\n",
    "MODE = 'min'\n",
    "VERBOSE = 1\n",
    "FACTOR = 0.1\n",
    "PATIENCE = 5\n",
    "COOLDOWN = 5\n",
    "BEST_MODEL = 'best_densenet.h5'\n",
    "FINAL_MODEL = 'best_densenet_final.h5'\n",
    "LOG_FILE = 'logs.csv'\n",
    "LOSS = 'binary_crossentropy'\n",
    "METRICS = ['accuracy', average_precision, f1_score]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train['Target'] = df_train['Target'].astype(str); \n",
    "df_valid['Target'] = df_valid['Target'].astype(str); \n",
    "df_test['Target'] = df_test['Target'].astype(str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Create generators for training, validation and test dataframes\n",
      "--------------------------------------------------------------------------------\n",
      "Train Generator Created ----------------------------------------\n",
      "Found 21348 validated image filenames belonging to 2 classes.\n",
      "Validation Generator Created ----------------------------------------\n",
      "Found 2668 validated image filenames belonging to 2 classes.\n",
      "Test Generator Created ----------------------------------------\n",
      "Found 2668 validated image filenames belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "print('Create generators for training, validation and test dataframes'); print('--'*40)\n",
    "generators = DataGenerators(df_train, df_valid, df_test, \n",
    "                            batch_size = BATCH_SIZE, \n",
    "                            path = TRAIN_IMAGES_DIR, \n",
    "                            img_size = (IMAGE_SIZE, IMAGE_SIZE), \n",
    "                            class_mode = 'binary',\n",
    "                            random_state = random_state)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "def build_model():\n",
    "    print('Create a `DenseNet121` model'); print('--'*40)\n",
    "    input_shape = (IMAGE_SIZE, IMAGE_SIZE, 3)\n",
    "    inputs = Input(shape = input_shape)\n",
    "    initializer = tf.keras.initializers.GlorotNormal()\n",
    "    \n",
    "    base_model = DenseNet121(include_top = False, input_tensor = inputs, weights ='imagenet')\n",
    "    \n",
    "    densenet = Model(inputs = inputs, outputs = base_model.layers[-1].output, name = 'DenseNet121')\n",
    "    \n",
    "    model = Sequential(name = 'DenseNet121')\n",
    "    model.add(densenet)\n",
    "    model.add(GlobalAveragePooling2D())\n",
    "    model.add(Dropout(0.4))\n",
    "    model.add(Dense(1, activation = 'sigmoid', kernel_initializer = initializer))\n",
    "    model.summary()\n",
    "    return model\n",
    "\n",
    "def callbacks():\n",
    "    lrscheduler = ReduceLROnPlateau(monitor = MONITOR, factor = FACTOR, \n",
    "                                    patience = PATIENCE, verbose = VERBOSE, \n",
    "                                    mode = MODE, cooldown = COOLDOWN)\n",
    "    \n",
    "    model.compile(optimizer = Adam(lr = LEARNING_RATE), loss = LOSS, metrics = METRICS)\n",
    "    \n",
    "    cp = ModelCheckpoint(filepath = MODEL_WEIGHTS + BEST_MODEL, monitor = MONITOR, \n",
    "                         verbose = VERBOSE, save_best_only = True, mode = MODE)\n",
    "    \n",
    "    if os.path.exists(MODEL_WEIGHTS + LOG_FILE): os.remove(MODEL_WEIGHTS + LOG_FILE)\n",
    "    csv_logger = CSVLogger(MODEL_WEIGHTS + LOG_FILE, append = True)\n",
    "    \n",
    "    callbacks = [cp, csv_logger, lrscheduler]\n",
    "    return callbacks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 394
    },
    "colab_type": "code",
    "id": "Ed9btNzeauai",
    "outputId": "642e9ca4-3607-4779-ccec-df0dc6a5ec4f",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lets fit the model.....\n",
      "Create a `DenseNet121` model\n",
      "--------------------------------------------------------------------------------\n",
      "Model: \"DenseNet121\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "DenseNet121 (Model)          (None, 7, 7, 1024)        7037504   \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d (Gl (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 1)                 1025      \n",
      "=================================================================\n",
      "Total params: 7,038,529\n",
      "Trainable params: 6,954,881\n",
      "Non-trainable params: 83,648\n",
      "_________________________________________________________________\n",
      "Train for 668 steps, validate for 84 steps\n",
      "Epoch 1/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.4231 - accuracy: 0.8103 - average_precision: 0.6078 - f1_score: 0.5038\n",
      "Epoch 00001: val_loss improved from inf to 0.50352, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_densenet.h5\n",
      "668/668 [==============================] - 445s 666ms/step - loss: 0.4230 - accuracy: 0.8104 - average_precision: 0.6078 - f1_score: 0.5040 - val_loss: 0.5035 - val_accuracy: 0.7774 - val_average_precision: 0.0952 - val_f1_score: 0.0324\n",
      "Epoch 2/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3729 - accuracy: 0.8353 - average_precision: 0.6895 - f1_score: 0.5650\n",
      "Epoch 00002: val_loss improved from 0.50352 to 0.45368, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_densenet.h5\n",
      "668/668 [==============================] - 415s 621ms/step - loss: 0.3727 - accuracy: 0.8354 - average_precision: 0.6894 - f1_score: 0.5651 - val_loss: 0.4537 - val_accuracy: 0.8141 - val_average_precision: 0.5664 - val_f1_score: 0.6074\n",
      "Epoch 3/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3608 - accuracy: 0.8411 - average_precision: 0.7034 - f1_score: 0.5841\n",
      "Epoch 00003: val_loss improved from 0.45368 to 0.34339, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_densenet.h5\n",
      "668/668 [==============================] - 412s 617ms/step - loss: 0.3605 - accuracy: 0.8413 - average_precision: 0.7039 - f1_score: 0.5846 - val_loss: 0.3434 - val_accuracy: 0.8478 - val_average_precision: 0.7373 - val_f1_score: 0.5698\n",
      "Epoch 4/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3524 - accuracy: 0.8428 - average_precision: 0.7071 - f1_score: 0.5884\n",
      "Epoch 00004: val_loss improved from 0.34339 to 0.33793, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_densenet.h5\n",
      "668/668 [==============================] - 412s 617ms/step - loss: 0.3525 - accuracy: 0.8427 - average_precision: 0.7070 - f1_score: 0.5881 - val_loss: 0.3379 - val_accuracy: 0.8519 - val_average_precision: 0.7128 - val_f1_score: 0.6193\n",
      "Epoch 5/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3455 - accuracy: 0.8453 - average_precision: 0.7222 - f1_score: 0.5905\n",
      "Epoch 00005: val_loss did not improve from 0.33793\n",
      "668/668 [==============================] - 411s 615ms/step - loss: 0.3454 - accuracy: 0.8454 - average_precision: 0.7226 - f1_score: 0.5905 - val_loss: 0.3431 - val_accuracy: 0.8478 - val_average_precision: 0.7625 - val_f1_score: 0.5670\n",
      "Epoch 6/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3393 - accuracy: 0.8500 - average_precision: 0.7293 - f1_score: 0.6109\n",
      "Epoch 00006: val_loss did not improve from 0.33793\n",
      "668/668 [==============================] - 409s 612ms/step - loss: 0.3392 - accuracy: 0.8501 - average_precision: 0.7294 - f1_score: 0.6111 - val_loss: 0.3716 - val_accuracy: 0.8482 - val_average_precision: 0.7606 - val_f1_score: 0.5665\n",
      "Epoch 7/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3343 - accuracy: 0.8499 - average_precision: 0.7205 - f1_score: 0.6115\n",
      "Epoch 00007: val_loss did not improve from 0.33793\n",
      "668/668 [==============================] - 427s 639ms/step - loss: 0.3345 - accuracy: 0.8498 - average_precision: 0.7204 - f1_score: 0.6114 - val_loss: 0.3389 - val_accuracy: 0.8512 - val_average_precision: 0.6805 - val_f1_score: 0.6330\n",
      "Epoch 8/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3294 - accuracy: 0.8515 - average_precision: 0.7266 - f1_score: 0.6143\n",
      "Epoch 00008: val_loss improved from 0.33793 to 0.33602, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_densenet.h5\n",
      "668/668 [==============================] - 411s 616ms/step - loss: 0.3293 - accuracy: 0.8516 - average_precision: 0.7267 - f1_score: 0.6144 - val_loss: 0.3360 - val_accuracy: 0.8448 - val_average_precision: 0.7009 - val_f1_score: 0.5996\n",
      "Epoch 9/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3238 - accuracy: 0.8559 - average_precision: 0.7406 - f1_score: 0.6273\n",
      "Epoch 00009: val_loss did not improve from 0.33602\n",
      "668/668 [==============================] - 411s 615ms/step - loss: 0.3238 - accuracy: 0.8559 - average_precision: 0.7400 - f1_score: 0.6270 - val_loss: 0.3621 - val_accuracy: 0.8497 - val_average_precision: 0.6729 - val_f1_score: 0.6255\n",
      "Epoch 10/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3169 - accuracy: 0.8584 - average_precision: 0.7415 - f1_score: 0.6358\n",
      "Epoch 00010: val_loss did not improve from 0.33602\n",
      "668/668 [==============================] - 410s 614ms/step - loss: 0.3170 - accuracy: 0.8584 - average_precision: 0.7419 - f1_score: 0.6357 - val_loss: 0.4284 - val_accuracy: 0.8430 - val_average_precision: 0.7776 - val_f1_score: 0.5379\n",
      "Save the final weights\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "print('Lets fit the model.....')\n",
    "K.clear_session()\n",
    "with strategy.scope():\n",
    "    model = build_model()\n",
    "    callbacks = callbacks()\n",
    "    train_generator = generators.train_generator\n",
    "    validation_generator = generators.valid_generator\n",
    "    test_generator = generators.test_generator    \n",
    "    history = model.fit_generator(generator = train_generator, \n",
    "                                  steps_per_epoch = generators.step_size_train,\n",
    "                                  epochs = EPOCH, verbose = VERBOSE, \n",
    "                                  callbacks = callbacks,\n",
    "                                  validation_data = validation_generator, \n",
    "                                  validation_steps = generators.step_size_valid)\n",
    "    print('Save the final weights'); print('--'*40)\n",
    "    model.save(MODEL_WEIGHTS + FINAL_MODEL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sbmJNma0J_ap"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reloading model weights and lets evaluate on validation set\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "print('Reloading model weights and lets evaluate on validation set'); print('--'*40)\n",
    "model.load_weights(MODEL_WEIGHTS + BEST_MODEL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "CG9dyWYMYZNk"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluate the model on validation data\n",
      "--------------------------------------------------------------------------------\n",
      "Loss: 0.336, Accuracy: 0.845, AP: 0.701, F1 Score: 0.6\n"
     ]
    }
   ],
   "source": [
    "print('Evaluate the model on validation data'); print('--'*40)\n",
    "\n",
    "loss, accuracy, ap, f1 = model.evaluate_generator(generator = validation_generator, \n",
    "                                          steps = generators.step_size_valid)\n",
    "print(f'Loss: {round(loss, 3)}, Accuracy: {round(float(accuracy), 3)}, AP: {round(float(ap), 3)}, F1 Score: {round(float(f1), 3)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "dF1DVIzFYmTG"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predict on the validation data\n",
      "--------------------------------------------------------------------------------\n",
      "84/84 [==============================] - 26s 313ms/step\n"
     ]
    }
   ],
   "source": [
    "print('Predict on the validation data'); print('--'*40)\n",
    "validation_generator.reset()\n",
    "valid_pred_roc = model.predict_generator(generator = validation_generator,\n",
    "                                         steps = generators.step_size_valid,\n",
    "                                         verbose = 1)\n",
    "valid_pred = []\n",
    "for i in valid_pred_roc:\n",
    "    if i >= 0.5: valid_pred.append(1)\n",
    "    else: valid_pred.append(0)\n",
    "y_valid = df_valid['Target'].astype(int).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "dCvXZQRYahnK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predict on the test data\n",
      "--------------------------------------------------------------------------------\n",
      "84/84 [==============================] - 26s 315ms/step\n"
     ]
    }
   ],
   "source": [
    "print('Predict on the test data'); print('--'*40)\n",
    "test_generator.reset()\n",
    "test_pred_roc = model.predict_generator(generator = test_generator,\n",
    "                                        steps = generators.step_size_test,\n",
    "                                        verbose = 1)\n",
    "test_pred = []\n",
    "for i in test_pred_roc:\n",
    "    if i >= 0.5: test_pred.append(1)\n",
    "    else: test_pred.append(0)\n",
    "y_test = df_test['Target'].astype(int).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    2071\n",
       "1     597\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "0    2042\n",
       "1     626\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(pd.Series(y_valid).value_counts(), pd.Series(y_test).value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ROC Curve for the validation data\n",
      "--------------------------------------------------------------------------------\n",
      "AUC: 0.890\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x518.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('ROC Curve for the validation data'); print('--'*40)\n",
    "\n",
    "roc_auc_valid = roc_auc_score(y_valid, np.array(valid_pred_roc).reshape(-1))\n",
    "print('AUC: {:0.3f}'.format(roc_auc_valid))\n",
    "\n",
    "fig = plt.figure(figsize = (10, 7.2))\n",
    "fpr, tpr, thresholds = roc_curve(y_valid, np.array(valid_pred_roc).reshape(-1))\n",
    "plt.title('ROC Curve for the validation data')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.axis([0, 1, 0, 1])\n",
    "plt.plot([0, 1], [0, 1], linestyle = '--', label = 'No Skill')\n",
    "plt.plot(fpr, tpr, marker = '.', label = 'ROC curve (area = %0.3f)' % roc_auc_valid)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ROC Curve for the test data\n",
      "--------------------------------------------------------------------------------\n",
      "AUC: 0.894\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x518.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('ROC Curve for the test data'); print('--'*40)\n",
    "\n",
    "roc_auc_test = roc_auc_score(y_test, np.array(test_pred_roc).reshape(-1))\n",
    "print('AUC: {:0.3f}'.format(roc_auc_test))\n",
    "\n",
    "fig = plt.figure(figsize = (10, 7.2))\n",
    "fpr, tpr, thresholds = roc_curve(y_test, np.array(test_pred_roc).reshape(-1))\n",
    "plt.title('ROC Curve for the test data')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.axis([0, 1, 0, 1])\n",
    "plt.plot([0, 1], [0, 1], linestyle = '--', label = 'Random')\n",
    "plt.plot(fpr, tpr, marker = '.', label = 'ROC curve (area = %0.3f)' % roc_auc_test)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix on the test data\n",
      "--------------------------------------------------------------------------------\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x518.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Confusion Matrix on the test data'); print('--'*40)\n",
    "cm = confusion_matrix(y_test, test_pred)\n",
    "plot_confusion_matrix(cm, figsize = (10, 7.2), cmap = plt.cm.Blues)\n",
    "plt.xticks(range(2), ['Normal', 'Pneumonia'], fontsize = 16)\n",
    "plt.yticks(range(2), ['Normal', 'Pneumonia'], fontsize = 16)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report on the test data\n",
      "------------------------------------------------------------------------------------------------------------------------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      Normal       0.87      0.93      0.90      2042\n",
      "   Pneumonia       0.72      0.55      0.62       626\n",
      "\n",
      "    accuracy                           0.84      2668\n",
      "   macro avg       0.79      0.74      0.76      2668\n",
      "weighted avg       0.84      0.84      0.84      2668\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('Classification Report on the test data'); print('--'*60)\n",
    "print(classification_report(y_test, test_pred, target_names = ['Normal', 'Pneumonia']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model - DenseNet121 + CheXNet pre-trained weights\n",
    "Acknowledgement for the pre-trained weights:\n",
    "We used pre-trained weights available from the following [repository](https://github.com/brucechou1983/CheXNet-Keras)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MVGCIGX6Jqqp"
   },
   "outputs": [],
   "source": [
    "# Model Parameters\n",
    "BATCH_SIZE = 32\n",
    "IMAGE_SIZE = 224\n",
    "EPOCH = 10\n",
    "LEARNING_RATE = 1e-3\n",
    "MONITOR = 'val_loss'\n",
    "MODE = 'min'\n",
    "VERBOSE = 1\n",
    "FACTOR = 0.1\n",
    "PATIENCE = 5\n",
    "COOLDOWN = 5\n",
    "BEST_MODEL = 'best_chexnet.h5'\n",
    "FINAL_MODEL = 'best_chexnet_final.h5'\n",
    "LOG_FILE = 'logs_chexnet.csv'\n",
    "LOSS = 'binary_crossentropy'\n",
    "METRICS = ['accuracy', average_precision, f1_score]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Create generators for training, validation and test dataframes\n",
      "--------------------------------------------------------------------------------\n",
      "Train Generator Created ----------------------------------------\n",
      "Found 21348 validated image filenames belonging to 2 classes.\n",
      "Validation Generator Created ----------------------------------------\n",
      "Found 2668 validated image filenames belonging to 2 classes.\n",
      "Test Generator Created ----------------------------------------\n",
      "Found 2668 validated image filenames belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "print('Create generators for training, validation and test dataframes'); print('--'*40)\n",
    "generators = DataGenerators(df_train, df_valid, df_test, \n",
    "                            batch_size = BATCH_SIZE, \n",
    "                            path = TRAIN_IMAGES_DIR, \n",
    "                            img_size = (IMAGE_SIZE, IMAGE_SIZE), \n",
    "                            class_mode = 'binary',\n",
    "                            random_state = random_state)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "def build_model():\n",
    "    print('Create a `CheXNet-like` model using pre-trained weights'); print('--'*40)\n",
    "    input_shape = (IMAGE_SIZE, IMAGE_SIZE, 3)\n",
    "    inputs = Input(shape = input_shape)\n",
    "    initializer = tf.keras.initializers.GlorotNormal()\n",
    "    \n",
    "    base_model = DenseNet121(include_top = True, input_tensor = inputs, weights = MODEL_WEIGHTS + 'CheXNet weights.h5', classes = 14)\n",
    "    \n",
    "    densenet = Model(inputs = inputs, outputs = base_model.layers[-2].output, name = 'CheXNet-like')\n",
    "    \n",
    "    model = Sequential(name = 'CheXNet-like')\n",
    "    model.add(densenet)\n",
    "    model.add(Dropout(0.2))\n",
    "    model.add(Dense(1, activation = 'sigmoid', kernel_initializer = initializer))\n",
    "    model.summary()\n",
    "    return model\n",
    "\n",
    "def callbacks():\n",
    "    lrscheduler = ReduceLROnPlateau(monitor = MONITOR, factor = FACTOR, \n",
    "                                    patience = PATIENCE, verbose = VERBOSE, \n",
    "                                    mode = MODE, cooldown = COOLDOWN)\n",
    "    \n",
    "    model.compile(optimizer = Adam(lr = LEARNING_RATE), loss = LOSS, metrics = METRICS)\n",
    "    \n",
    "    cp = ModelCheckpoint(filepath = MODEL_WEIGHTS + BEST_MODEL, monitor = MONITOR, \n",
    "                         verbose = VERBOSE, save_best_only = True, mode = MODE)\n",
    "    \n",
    "    if os.path.exists(MODEL_WEIGHTS + LOG_FILE): os.remove(MODEL_WEIGHTS + LOG_FILE)\n",
    "    csv_logger = CSVLogger(MODEL_WEIGHTS + LOG_FILE, append = True)\n",
    "    \n",
    "    callbacks = [cp, csv_logger, lrscheduler]\n",
    "    return callbacks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 394
    },
    "colab_type": "code",
    "id": "Ed9btNzeauai",
    "outputId": "642e9ca4-3607-4779-ccec-df0dc6a5ec4f",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lets fit the model.....\n",
      "Create a `CheXNet-like` model using pre-trained weights\n",
      "--------------------------------------------------------------------------------\n",
      "Model: \"CheXNet-like\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "CheXNet-like (Model)         (None, 1024)              7037504   \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 1)                 1025      \n",
      "=================================================================\n",
      "Total params: 7,038,529\n",
      "Trainable params: 6,954,881\n",
      "Non-trainable params: 83,648\n",
      "_________________________________________________________________\n",
      "Train for 668 steps, validate for 84 steps\n",
      "Epoch 1/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3733 - accuracy: 0.8300 - average_precision: 0.6600 - f1_score: 0.5311\n",
      "Epoch 00001: val_loss improved from inf to 0.44532, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_chexnet.h5\n",
      "668/668 [==============================] - 419s 627ms/step - loss: 0.3732 - accuracy: 0.8301 - average_precision: 0.6605 - f1_score: 0.5312 - val_loss: 0.4453 - val_accuracy: 0.8055 - val_average_precision: 0.6056 - val_f1_score: 0.3313\n",
      "Epoch 2/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3533 - accuracy: 0.8403 - average_precision: 0.7008 - f1_score: 0.5837\n",
      "Epoch 00002: val_loss improved from 0.44532 to 0.37923, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_chexnet.h5\n",
      "668/668 [==============================] - 405s 606ms/step - loss: 0.3533 - accuracy: 0.8404 - average_precision: 0.7008 - f1_score: 0.5838 - val_loss: 0.3792 - val_accuracy: 0.8332 - val_average_precision: 0.7764 - val_f1_score: 0.4595\n",
      "Epoch 3/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3511 - accuracy: 0.8427 - average_precision: 0.7070 - f1_score: 0.5909\n",
      "Epoch 00003: val_loss did not improve from 0.37923\n",
      "668/668 [==============================] - 404s 605ms/step - loss: 0.3511 - accuracy: 0.8427 - average_precision: 0.7075 - f1_score: 0.5911 - val_loss: 0.4044 - val_accuracy: 0.8388 - val_average_precision: 0.7769 - val_f1_score: 0.4942\n",
      "Epoch 4/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3450 - accuracy: 0.8426 - average_precision: 0.7085 - f1_score: 0.5871\n",
      "Epoch 00004: val_loss did not improve from 0.37923\n",
      "668/668 [==============================] - 404s 604ms/step - loss: 0.3449 - accuracy: 0.8427 - average_precision: 0.7089 - f1_score: 0.5874 - val_loss: 0.4144 - val_accuracy: 0.8306 - val_average_precision: 0.8101 - val_f1_score: 0.4419\n",
      "Epoch 5/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3449 - accuracy: 0.8457 - average_precision: 0.7130 - f1_score: 0.5984\n",
      "Epoch 00005: val_loss did not improve from 0.37923\n",
      "668/668 [==============================] - 404s 604ms/step - loss: 0.3450 - accuracy: 0.8458 - average_precision: 0.7135 - f1_score: 0.5986 - val_loss: 1.1231 - val_accuracy: 0.7980 - val_average_precision: 0.5258 - val_f1_score: 0.1850\n",
      "Epoch 6/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3409 - accuracy: 0.8464 - average_precision: 0.7148 - f1_score: 0.6002\n",
      "Epoch 00006: val_loss did not improve from 0.37923\n",
      "668/668 [==============================] - 403s 604ms/step - loss: 0.3410 - accuracy: 0.8464 - average_precision: 0.7143 - f1_score: 0.5999 - val_loss: 0.4552 - val_accuracy: 0.8370 - val_average_precision: 0.7718 - val_f1_score: 0.5021\n",
      "Epoch 7/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3380 - accuracy: 0.8475 - average_precision: 0.7116 - f1_score: 0.6055\n",
      "Epoch 00007: val_loss did not improve from 0.37923\n",
      "\n",
      "Epoch 00007: ReduceLROnPlateau reducing learning rate to 0.00010000000474974513.\n",
      "668/668 [==============================] - 403s 604ms/step - loss: 0.3379 - accuracy: 0.8475 - average_precision: 0.7117 - f1_score: 0.6056 - val_loss: 1.3599 - val_accuracy: 0.7777 - val_average_precision: 0.0357 - val_f1_score: 0.0119\n",
      "Epoch 8/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3215 - accuracy: 0.8585 - average_precision: 0.7409 - f1_score: 0.6375\n",
      "Epoch 00008: val_loss improved from 0.37923 to 0.32581, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_chexnet.h5\n",
      "668/668 [==============================] - 404s 605ms/step - loss: 0.3213 - accuracy: 0.8587 - average_precision: 0.7413 - f1_score: 0.6379 - val_loss: 0.3258 - val_accuracy: 0.8538 - val_average_precision: 0.7211 - val_f1_score: 0.6140\n",
      "Epoch 9/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3172 - accuracy: 0.8566 - average_precision: 0.7317 - f1_score: 0.6339\n",
      "Epoch 00009: val_loss improved from 0.32581 to 0.32542, saving model to /home/pratik_sharma_ds/Capstone_Pneumonia_Detection/model_weights/best_chexnet.h5\n",
      "668/668 [==============================] - 408s 611ms/step - loss: 0.3170 - accuracy: 0.8568 - average_precision: 0.7316 - f1_score: 0.6341 - val_loss: 0.3254 - val_accuracy: 0.8482 - val_average_precision: 0.6918 - val_f1_score: 0.6097\n",
      "Epoch 10/10\n",
      "667/668 [============================>.] - ETA: 0s - loss: 0.3083 - accuracy: 0.8623 - average_precision: 0.7464 - f1_score: 0.6457\n",
      "Epoch 00010: val_loss did not improve from 0.32542\n",
      "668/668 [==============================] - 405s 606ms/step - loss: 0.3085 - accuracy: 0.8621 - average_precision: 0.7465 - f1_score: 0.6455 - val_loss: 0.3320 - val_accuracy: 0.8482 - val_average_precision: 0.6931 - val_f1_score: 0.6104\n",
      "Save the final weights\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "print('Lets fit the model.....')\n",
    "K.clear_session()\n",
    "with strategy.scope():\n",
    "    model = build_model()\n",
    "    callbacks = callbacks()\n",
    "    train_generator = generators.train_generator\n",
    "    validation_generator = generators.valid_generator\n",
    "    test_generator = generators.test_generator    \n",
    "    history = model.fit_generator(generator = train_generator, \n",
    "                                  steps_per_epoch = generators.step_size_train,\n",
    "                                  epochs = EPOCH, verbose = VERBOSE, \n",
    "                                  callbacks = callbacks,\n",
    "                                  validation_data = validation_generator, \n",
    "                                  validation_steps = generators.step_size_valid)\n",
    "    print('Save the final weights'); print('--'*40)\n",
    "    model.save(MODEL_WEIGHTS + FINAL_MODEL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sbmJNma0J_ap",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reloading model weights and lets evaluate on validation set\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "print('Reloading model weights and lets evaluate on validation set'); print('--'*40)\n",
    "model.load_weights(MODEL_WEIGHTS + BEST_MODEL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "CG9dyWYMYZNk"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluate the model on validation data\n",
      "--------------------------------------------------------------------------------\n",
      "Loss: 0.325, Accuracy: 0.848, AP: 0.692, F1 Score: 0.61\n"
     ]
    }
   ],
   "source": [
    "print('Evaluate the model on validation data'); print('--'*40)\n",
    "\n",
    "loss, accuracy, ap, f1 = model.evaluate_generator(generator = validation_generator, \n",
    "                                          steps = generators.step_size_valid)\n",
    "print(f'Loss: {round(loss, 3)}, Accuracy: {round(float(accuracy), 3)}, AP: {round(float(ap), 3)}, F1 Score: {round(float(f1), 3)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "dF1DVIzFYmTG"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predict on the validation data\n",
      "--------------------------------------------------------------------------------\n",
      "84/84 [==============================] - 26s 314ms/step\n"
     ]
    }
   ],
   "source": [
    "print('Predict on the validation data'); print('--'*40)\n",
    "validation_generator.reset()\n",
    "valid_pred_roc = model.predict_generator(generator = validation_generator,\n",
    "                                         steps = generators.step_size_valid,\n",
    "                                         verbose = 1)\n",
    "valid_pred = []\n",
    "for i in valid_pred_roc:\n",
    "    if i >= 0.5: valid_pred.append(1)\n",
    "    else: valid_pred.append(0)\n",
    "y_valid = df_valid['Target'].astype(int).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "dCvXZQRYahnK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predict on the test data\n",
      "--------------------------------------------------------------------------------\n",
      "84/84 [==============================] - 26s 311ms/step\n"
     ]
    }
   ],
   "source": [
    "print('Predict on the test data'); print('--'*40)\n",
    "test_generator.reset()\n",
    "test_pred_roc = model.predict_generator(generator = test_generator,\n",
    "                                        steps = generators.step_size_test,\n",
    "                                        verbose = 1)\n",
    "test_pred = []\n",
    "for i in test_pred_roc:\n",
    "    if i >= 0.5: test_pred.append(1)\n",
    "    else: test_pred.append(0)\n",
    "y_test = df_test['Target'].astype(int).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    2071\n",
       "1     597\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "0    2042\n",
       "1     626\n",
       "dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(pd.Series(y_valid).value_counts(), pd.Series(y_test).value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ROC Curve for the validation data\n",
      "--------------------------------------------------------------------------------\n",
      "AUC: 0.898\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x518.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('ROC Curve for the validation data'); print('--'*40)\n",
    "\n",
    "roc_auc_valid = roc_auc_score(y_valid, np.array(valid_pred_roc).reshape(-1))\n",
    "print('AUC: {:0.3f}'.format(roc_auc_valid))\n",
    "\n",
    "fig = plt.figure(figsize = (10, 7.2))\n",
    "fpr, tpr, thresholds = roc_curve(y_valid, np.array(valid_pred_roc).reshape(-1))\n",
    "plt.title('ROC Curve for the validation data')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.axis([0, 1, 0, 1])\n",
    "plt.plot([0, 1], [0, 1], linestyle = '--', label = 'No Skill')\n",
    "plt.plot(fpr, tpr, marker = '.', label = 'ROC curve (area = %0.3f)' % roc_auc_valid)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ROC Curve for the test data\n",
      "--------------------------------------------------------------------------------\n",
      "AUC: 0.895\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x518.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('ROC Curve for the test data'); print('--'*40)\n",
    "\n",
    "roc_auc_test = roc_auc_score(y_test, np.array(test_pred_roc).reshape(-1))\n",
    "print('AUC: {:0.3f}'.format(roc_auc_test))\n",
    "\n",
    "fig = plt.figure(figsize = (10, 7.2))\n",
    "fpr, tpr, thresholds = roc_curve(y_test, np.array(test_pred_roc).reshape(-1))\n",
    "plt.title('ROC Curve for the test data')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.axis([0, 1, 0, 1])\n",
    "plt.plot([0, 1], [0, 1], linestyle = '--', label = 'Random')\n",
    "plt.plot(fpr, tpr, marker = '.', label = 'ROC curve (area = %0.3f)' % roc_auc_test)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix on the test data\n",
      "--------------------------------------------------------------------------------\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x518.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Confusion Matrix on the test data'); print('--'*40)\n",
    "cm = confusion_matrix(y_test, test_pred)\n",
    "plot_confusion_matrix(cm, figsize = (10, 7.2), cmap = plt.cm.Blues)\n",
    "plt.xticks(range(2), ['Normal', 'Pneumonia'], fontsize = 16)\n",
    "plt.yticks(range(2), ['Normal', 'Pneumonia'], fontsize = 16)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classification Report on the test data\n",
      "------------------------------------------------------------------------------------------------------------------------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "      Normal       0.87      0.93      0.90      2042\n",
      "   Pneumonia       0.72      0.55      0.63       626\n",
      "\n",
      "    accuracy                           0.85      2668\n",
      "   macro avg       0.80      0.74      0.76      2668\n",
      "weighted avg       0.84      0.85      0.84      2668\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('Classification Report on the test data'); print('--'*60)\n",
    "print(classification_report(y_test, test_pred, target_names = ['Normal', 'Pneumonia']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train/Validation Accuracy/Loss Comparison\n",
      "--------------------------------------------------------------------------------\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print('Train/Validation Accuracy/Loss Comparison'); print('--'*40)\n",
    "logs = pd.read_csv(MODEL_WEIGHTS + 'logs.csv')\n",
    "logs_chex = pd.read_csv(MODEL_WEIGHTS + 'logs_chexnet.csv')\n",
    "f, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, figsize = (12, 8))\n",
    "\n",
    "ax1.plot(logs['epoch'], logs['accuracy'], 'g', label = 'DenseNet121 Training Accuracy')\n",
    "ax1.plot(logs_chex['epoch'], logs_chex['accuracy'], 'b', label = 'CheXNet Training Accuracy')\n",
    "ax1.set_title('Training Accuracy for DenseNet121 and CheXNet')\n",
    "ax1.set_xlabel('Epochs')\n",
    "ax1.set_ylabel('Accuracy')\n",
    "ax1.set_xlim(0, logs['epoch'].max())\n",
    "# ax1.set_ylim(0, 1)\n",
    "ax1.legend(loc = 'lower right')\n",
    "\n",
    "ax2.plot(logs['epoch'], logs['val_accuracy'], 'g', label = 'DenseNet121 Validation Accuracy')\n",
    "ax2.plot(logs_chex['epoch'], logs_chex['val_accuracy'], 'b', label = 'CheXNet Validation Accuracy')\n",
    "ax2.set_title('Validation Accuracy for DenseNet121 and CheXNet')\n",
    "ax2.set_xlabel('Epochs')\n",
    "ax2.set_ylabel('Accuracy')\n",
    "ax2.set_xlim(0, logs['epoch'].max())\n",
    "# ax2.set_ylim(0, 1)\n",
    "ax2.legend(loc = 'lower right')\n",
    "\n",
    "ax3.plot(logs['epoch'], logs['loss'], 'g', label = 'DenseNet121 Training Loss')\n",
    "ax3.plot(logs_chex['epoch'], logs_chex['loss'], 'b', label = 'CheXNet Training Loss')\n",
    "ax3.set_title('Training Loss for DenseNet121 and CheXNet')\n",
    "ax3.set_xlabel('Epochs')\n",
    "ax3.set_ylabel('Loss')\n",
    "ax3.set_xlim(0, logs['epoch'].max())\n",
    "# ax3.set_ylim(0, logs['loss'].max())\n",
    "ax3.legend(loc = 'lower left')\n",
    "\n",
    "ax4.plot(logs['epoch'], logs['val_loss'], 'g', label = 'DenseNet121 Validation Loss')\n",
    "ax4.plot(logs_chex['epoch'], logs_chex['val_loss'], 'b', label = 'CheXNet Validation Loss')\n",
    "ax4.set_title('Validation Loss for DenseNet121 and CheXNet')\n",
    "ax4.set_xlabel('Epochs')\n",
    "ax4.set_ylabel('Loss')\n",
    "ax4.set_xlim(0, logs['epoch'].max())\n",
    "# ax4.set_ylim(0, logs['val_loss'].max())\n",
    "ax4.legend(loc = 'lower left')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Pneumonia_Detection_Model_DenseNet121.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}