covidvspne.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_dGgVb8DPZMp",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# importing neccessary libraries\n",
    "from glob import glob\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import random\n",
    "from skimage.io import imread\n",
    "import os, shutil\n",
    "import tensorflow as tf\n",
    "import keras\n",
    "from tensorflow.keras.layers import Input, MaxPooling2D\n",
    "from collections import Counter\n",
    "from scikitplot.metrics import plot_roc\n",
    "from sklearn import metrics\n",
    "import matplotlib.image as mpimg\n",
    "import matplotlib.pyplot as plt\n",
    "from PIL import Image\n",
    "import seaborn as sns\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "import scikitplot\n",
    "import sklearn\n",
    "import cv2\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from sklearn.linear_model import Ridge\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.preprocessing import MinMaxScaler, StandardScaler\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.ensemble import VotingClassifier,AdaBoostClassifier,RandomForestClassifier,ExtraTreesClassifier,GradientBoostingClassifier,GradientBoostingRegressor\n",
    "from sklearn.tree import DecisionTreeClassifier, export_graphviz\n",
    "from sklearn.feature_selection import VarianceThreshold\n",
    "from sklearn import datasets \n",
    "from tensorflow.keras.applications import VGG16\n",
    "from tensorflow.keras.layers import AveragePooling2D\n",
    "from sklearn import tree\n",
    "from sklearn.naive_bayes import GaussianNB \n",
    "from sklearn.model_selection import validation_curve\n",
    "import imblearn\n",
    "from sklearn.decomposition import TruncatedSVD\n",
    "from sklearn.metrics import precision_score, precision_recall_curve, average_precision_score, recall_score, accuracy_score, f1_score, make_scorer, classification_report, auc, roc_curve, roc_auc_score, confusion_matrix\n",
    "from imblearn.pipeline import make_pipeline as make_pipeline_imb\n",
    "from imblearn.under_sampling import NearMiss\n",
    "from imblearn.metrics import classification_report_imbalanced\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from keras.applications import MobileNet, DenseNet201, DenseNet121, VGG19, InceptionV3, MobileNetV2, Xception, ResNet152V2, NASNetLarge\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten, GlobalAveragePooling2D\n",
    "from keras.layers import Conv2D, MaxPool2D, ZeroPadding2D\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.applications.resnet50 import preprocess_input, decode_predictions\n",
    "from tensorflow.python.keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img\n",
    "from keras.optimizers import Adam, SGD\n",
    "from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Processessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 340
    },
    "colab_type": "code",
    "id": "6pw0aLmIeoe0",
    "outputId": "32f6f93a-4837-49b5-f609-49000c28a1f6",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total\n",
      "---------------------\n",
      "pneumonia no train:  845\n",
      "pneumonia no val:  226\n",
      "pneumonia no test:  270\n",
      "---------------------\n",
      "covid no train:  666\n",
      "covid no val:  167\n",
      "covid no test:  39\n",
      "\n",
      "Total\n",
      "---------------------\n",
      "total train:  1511\n",
      "total val:  393\n",
      "total test:  309\n"
     ]
    }
   ],
   "source": [
    "# counting the amount of x rays images\n",
    "len_pneumonia_train = len([iq for iq in os.scandir('xrays/covidvspneumonia/train/pneumonia')])\n",
    "len_pneumonia_val = len([iq for iq in os.scandir('xrays/covidvspneumonia/val/pneumonia')])\n",
    "len_pneumonia_test = len([iq for iq in os.scandir('xrays/covidvspneumonia/test/pneumonia')])\n",
    "\n",
    "len_covid_train = len([iq for iq in os.scandir('xrays/covidvspneumonia/train/covid')])\n",
    "len_covid_val = len([iq for iq in os.scandir('xrays/covidvspneumonia/val/covid')])\n",
    "len_covid_test = len([iq for iq in os.scandir('xrays/covidvspneumonia/test/covid')])\n",
    "\n",
    "\n",
    "len_train_total = len_covid_train + len_pneumonia_train\n",
    "len_val_total = len_pneumonia_val + len_covid_val\n",
    "len_test_total = len_covid_test + len_pneumonia_test\n",
    "\n",
    "print(\"Total\")\n",
    "print(\"---------------------\")\n",
    "print (\"pneumonia no train: \", len_pneumonia_train)\n",
    "print (\"pneumonia no val: \", len_pneumonia_val)\n",
    "print (\"pneumonia no test: \", len_pneumonia_test)\n",
    "print(\"---------------------\")\n",
    "print (\"covid no train: \", len_covid_train)\n",
    "print (\"covid no val: \", len_covid_val)\n",
    "print (\"covid no test: \", len_covid_test)\n",
    "print()\n",
    "print(\"Total\")\n",
    "print(\"---------------------\")\n",
    "print(\"total train: \", len_train_total)\n",
    "print(\"total val: \", len_val_total)\n",
    "print(\"total test: \", len_test_total)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# extracting the images\n",
    "DIR_NAME = 'xrays/covidvspneumonia/'\n",
    "\n",
    "imagePaths=[]\n",
    "for dirname, _, filenames in os.walk(DIR_NAME):\n",
    "    for filename in filenames:\n",
    "        imagePaths.append(os.path.join(dirname, filename))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(152).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(153).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(154).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(155).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(156).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(157).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(158).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(159).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(160).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(161).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(162).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(163).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(164).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(165).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(166).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(167).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(168).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(169).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(170).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(171).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(172).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(173).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(174).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(175).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(176).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(177).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(178).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(179).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(180).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(181).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(182).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(183).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(184).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(185).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(186).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(187).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(188).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(189).png',\n",
       " 'xrays/covidvspneumonia/test\\\\covid\\\\COVID-19(190).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (538)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (539).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (539)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (540).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (540)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (541).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (541)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (542).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (542)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (543).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (543)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (544).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (544)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (545).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (545)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (546).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (546)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (547).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (547)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (548).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (548)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (549).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (549)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (550).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (550)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (551).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (551)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (552).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (552)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (553).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (553)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (554).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (554)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (555).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (555)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (556).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (556)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (557).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (557)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (558).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (558)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (559).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (559)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (560).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (560)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (561).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (561)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (562).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (562)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (563).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (563)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (564).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (564)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (565).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (565)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (566).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (566)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (567).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (567)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (568).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (568)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (569).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (569)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (570).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (570)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (571).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (571)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (572).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (572)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (573).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (573)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (574).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (574)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (575).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (575)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (576).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (576)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (577).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (577)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (578).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (578)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (579).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (579)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (580).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (580)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (581).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (581)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (582).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (582)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (583).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (583)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (584).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (584)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (585).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (585)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (586).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (586)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (587).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (587)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (588).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (588)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (589).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (589)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (590).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (590)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (591).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (591)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (592).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (592)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (593).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (593)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (594).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (594)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (595).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (595)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (596).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (596)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (597).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (597)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (598).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (598)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (599).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (599)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (600).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (600)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (601).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (601)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (602).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (602)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (603).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (603)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (604).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (604)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (605).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (605)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (606).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (606)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (607).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (607)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (608).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (608)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (609).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (609)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (610).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (610)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (611).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (611)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (612).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (612)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (613).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (613)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (614).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (614)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (615).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (615)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (616).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (616)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (617).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (617)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (618).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (618)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (619).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (619)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (620).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (620)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (621).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (621)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (622).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (622)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (623).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (623)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (624).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (624)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (625).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (625)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (626).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (626)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (627).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (627)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (628).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (628)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (629).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (629)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (630).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (630)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (631).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (631)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (632).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (632)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (633).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (633)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (634).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (634)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (635).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (635)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (636).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (636)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (637).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (637)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (638).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (638)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (639).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (639)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (640).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (640)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (641).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (641)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (642).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (642)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (643).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (643)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (644).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (644)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (645).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (645)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (646).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (646)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (647).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (647)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (648).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (648)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (649).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (649)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (650).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (650)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (651).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (651)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (652).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (652)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (653).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (653)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (654).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (654)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (655).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (655)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (656).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (656)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (657).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (657)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (658).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (658)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (659).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (659)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (660).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (660)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (661).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (661)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (662).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (662)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (663).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (663)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (664).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (664)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (665).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (665)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (666).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (666)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (667).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (667)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (668).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (668)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (669).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (669)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (670).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (670)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (671).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (671)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (672).png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (672)_1.png',\n",
       " 'xrays/covidvspneumonia/test\\\\pneumonia\\\\Viral Pneumonia (673).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\01E392EE-69F9-4E33-BFCE-E5C968654078.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\03BF7561-A9BA-4C3C-B8A0-D3E585F73F3C.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\0a7faa2a.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1-s2.0-S0140673620303706-fx1_lrg.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1-s2.0-S1684118220300608-main.pdf-001.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1-s2.0-S1684118220300682-main.pdf-002-a2.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1-s2.0-S1684118220300682-main.pdf-003-b1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1-s2.0-S1684118220300682-main.pdf-003-b2.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1.CXRCTThoraximagesofCOVID-19fromSingapore.pdf-000-fig1b.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1.CXRCTThoraximagesofCOVID-19fromSingapore.pdf-001-fig2a.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1.CXRCTThoraximagesofCOVID-19fromSingapore.pdf-002-fig3b.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\10.1016-slash-j.crad.2020.04.002-a.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\10.1016-slash-j.crad.2020.04.002-b.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\12941_2020_358_Fig1_HTML.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1312A392-67A3-4EBF-9319-810CF6DA5EF6.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16654_1_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16654_2_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16654_4_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16660_1_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16660_2_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16660_4_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16663_1_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16664_1_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16669_1_1.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16672_1_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16708_1_1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\16745_5_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\19abe1f3.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\1B734A89-A1BF-49A8-A1D3-66FAFA4FAC5D.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\23E99E2E-447C-46E5-8EB2-D35D12473C39.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\2C26F453-AF3B-4517-BB9E-802CF2179543.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\2cd63b76.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\31BA3780-2323-493F-8AED-62081B9C383B.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\353889E0-A1E8-4F9E-A0B8-F24F36BCFBFB.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\3b46de94.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\41182_2020_203_Fig3_HTML.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\41182_2020_203_Fig4_HTML.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\41182_2020_203_Fig5_HTML.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\41379_2020_536_Fig2_HTML-c1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\41591_2020_819_Fig1_HTML.webp-day10.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\41591_2020_819_Fig1_HTML.webp-day5.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\48c4542c.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\4d98e1de.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\4e43e48d52c9e2d4c6c1fb9bc1544f_jumbo.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\53EC07C9-5CC6-4BE4-9B6F-D7B0D72AAA7E.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\58cb9263f16e94305c730685358e4e_jumbo.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\6b3bdbc31f65230b8cdcc3cef5f8ba8a-40ac-0.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\6b44464d-73a7-4cf3-bbb6-ffe7168300e3.annot.original.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\6CB4EFC6-68FA-4CD5-940C-BEFA8DAFE9A7.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\7-fatal-covid19.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\7AF6C1AF-D249-4BD2-8C26-449304105D03.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\7afaaccc.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\7C69C012-7479-493F-8722-ABC29C60A2DD.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\80446565-E090-4187-A031-9D3CEAA586C8.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\85E52EB3-56E9-4D67-82DA-DEA247C82886.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\88de9d8c39e946abd495b37cd07d89e5-6531-0.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\8da14fff.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\8FDE8DBA-CFBD-4B4C-B1A4-6F36A93B7E87.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\93FE0BB1-022D-4F24-9727-987A07975FFB.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\9fdd3c3032296fd04d2cad5d9070d4_jumbo.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ac4f6e4e.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\all14238-fig-0001-m-b.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\all14238-fig-0001-m-c.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\AR-1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\AR-2.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\auntminnie-a-2020_01_28_23_51_6665_2020_01_28_Vietnam_coronavirus.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\auntminnie-c-2020_01_28_23_51_6665_2020_01_28_Vietnam_coronavirus.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\B2D20576-00B7-4519-A415-72DE29C90C34.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\b9673e89.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\bd10d5e2.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\BMJ-37-163-g1.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\c0f74558.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\CD50BA96-6982-4C80-AE7B-5F67ACDBFA56.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ciaa199.pdf-001-b.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ciaa199.pdf-001-c.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (1).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (10).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (109).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (11).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (112).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (113).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (114).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (115).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (118).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (119).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (12).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (120).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (121).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (122).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (123).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (124).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (125).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (126).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (127).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (128).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (129).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (13).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (130).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (131).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (132).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (133).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (14).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (15).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (16).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (17).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (18).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (19).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (2).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (20).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (21).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (22).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (26).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (28).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (29).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (3).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (31).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (33).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (36).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (37).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (39).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (4).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (42).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (43).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (44).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (45).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (46).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (48).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (49).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (5).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (50).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (51).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (52).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (53).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (54).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (56).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (58).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (59).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (6).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (61).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (63).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (64).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (65).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (66).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (67).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (68).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (69).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (7).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (70).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (71).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (72).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (73).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (74).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (75).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (76).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (77).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (78).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (79).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (8).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (80).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (81).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (82).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (85).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (86).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (87).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (88).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (89).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (9).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (90).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (91).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (94).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (95).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (96).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (97).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (98).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19 (99).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(134).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(136).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(137).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(138).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(139).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(140).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(145).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(146).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(147).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(148).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(149).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(150).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID-19(151).png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-caso-70-1-PA.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-caso-91-1-12.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-infection-exclusive-gastrointestinal-symptoms-pa.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-12.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-14-PA.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-15-PA.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-19.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-2.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-20.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-22-day2-pa.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-30-PA.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-34.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-35-2.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-38-pa.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-44-day-0.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-58-day-3.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-58-day-7.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-58-day-9.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-67.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-7-PA.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-93.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-bilateral.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\covid-19-pneumonia-evolution-over-a-week-1-day0-PA.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(0).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(1).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(10).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(100).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(101).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(102).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(103).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(104).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(105).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(106).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(107).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(108).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(109).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(11).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(110).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(111).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(112).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(113).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(114).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(115).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(116).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(117).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(119).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(12).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(120).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(121).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(122).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(123).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(124).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(125).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(126).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(127).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(128).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(129).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(13).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(130).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(131).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(132).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(133).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(134).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(135).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(136).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(137).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(138).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(139).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(14).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(140).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(141).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(142).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(143).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(144).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(145).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(146).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(147).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(148).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(149).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(15).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(150).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(151).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(152).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(153).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(154).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(155).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(156).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(157).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(158).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(159).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(160).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(161).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(162).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(163).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(164).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(165).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(166).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(167).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(168).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(169).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(17).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(170).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(171).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(172).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(173).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(174).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(175).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(176).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(177).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(178).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(179).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(18).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(180).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(181).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(182).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(183).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(184).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(185).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(186).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(187).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(188).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(189).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(19).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(190).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(191).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(192).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(193).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(194).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(195).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(196).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(197).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(198).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(199).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(2).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(20).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(200).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(201).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(202).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(203).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(204).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(205).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(206).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(207).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(208).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(209).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(21).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(210).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(211).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(212).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(213).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(214).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(215).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(216).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(217).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(218).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(219).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(220).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(221).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(222).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(223).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(224).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(225).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(226).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(227).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(228).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(229).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(230).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(231).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(232).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(234).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(235).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(236).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(237).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(238).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(239).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(24).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(240).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(241).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(242).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(243).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(244).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(245).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(246).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(247).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(248).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(249).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(25).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(250).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(251).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(252).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(253).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(254).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(255).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(256).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(257).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(258).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(26).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(260).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(261).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(262).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(263).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(264).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(265).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(266).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(267).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(268).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(269).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(27).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(270).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(271).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(272).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(273).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(274).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(275).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(276).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(277).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(278).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(279).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(28).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(280).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(281).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(282).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(283).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(284).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(285).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(286).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(287).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(288).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(289).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(29).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(290).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(291).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(292).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(293).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(294).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(295).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(296).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(297).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(298).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(299).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(3).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(30).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(309).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(31).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(313).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(314).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(315).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(316).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(317).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(318).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(319).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(32).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(320).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(321).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(322).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(323).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(324).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(325).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(326).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(327).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(328).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(329).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(33).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(330).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(331).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(332).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(333).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(334).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(335).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(336).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(337).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(338).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(339).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(34).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(340).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(341).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(342).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(343).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(344).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(345).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(346).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(347).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(348).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(349).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(35).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(350).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(351).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(352).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(353).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(354).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(355).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(356).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(357).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(358).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(359).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(36).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(360).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(361).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(362).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(363).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(364).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(365).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(366).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(367).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(368).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(369).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(37).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(370).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(371).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(374).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(375).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(376).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(377).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(378).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(379).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(38).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(380).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(381).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(382).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(383).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(384).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(385).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(386).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(387).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(388).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(389).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(39).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(390).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(391).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(392).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(393).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(394).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(395).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(396).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(397).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(398).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(399).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(4).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(40).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(400).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(401).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(402).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(403).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(404).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(405).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(406).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(407).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(408).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(409).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(41).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(410).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(411).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(412).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(413).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(414).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(415).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(416).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(417).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(418).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(419).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(42).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(420).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(421).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(422).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(423).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(424).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(425).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(426).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(427).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(428).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(429).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(43).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(430).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(431).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(432).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(433).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(434).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(435).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(436).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(437).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(438).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(439).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(44).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(440).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(441).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(442).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(443).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(444).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(445).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(446).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(447).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(448).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(449).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(45).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(450).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(451).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(452).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(453).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(454).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(455).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(456).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(457).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(458).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(459).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(46).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(47).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(48).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(49).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(5).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(50).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(51).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(52).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(53).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(54).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(57).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(58).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(59).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(6).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(60).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(61).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(62).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(63).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(64).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(65).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(7).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(73).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(74).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(75).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(76).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(77).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(78).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(79).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(8).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(80).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(81).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(82).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(83).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(84).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(85).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(86).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(87).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(88).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(89).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(9).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(90).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(91).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(93).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(94).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(95).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(96).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(97).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(98).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\COVID19(99).jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\d009d61f.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\d9d6ca9a.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ec083e35.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\F4341CE7-73C9-45C6-99C8-8567A5484B63.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\F63AB6CE-1968-4154-A70F-913AF154F53D.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\f64132c8.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\f6575117.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\f6d980a0.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\f8335316.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\FE9F9A5D-2830-46F9-851B-1FF4534959BE.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\fff49165-b22d-4bb4-b9d1-d5d62c52436c.annot.original.png',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\figure1-5e71be566aa8714a04de3386-98-left.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\figure1-5e75d0940b71e1b702629659-98-right.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\kjr-21-e24-g002-l-a.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\kjr-21-e25-g001-l-a.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\lancet-case2a.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\lancet-case2b.jpg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\nCoV-radiol.2020200269.fig1-day7.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\nejmc2001573_f1a.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\nejmc2001573_f1b.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\nejmoa2001191_f3-PA.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\nejmoa2001191_f4.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ryct.2020200034.fig2.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ryct.2020200034.fig5-day0.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\covid\\\\ryct.2020200034.fig5-day4.jpeg',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (1).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (1)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (10).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (10)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (100).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (100)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (101).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (101)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (102).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (102)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (103).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (103)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (104).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (104)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (105).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (105)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (106).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (106)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (107).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (107)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (108).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (108)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (109).png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (109)_1.png',\n",
       " 'xrays/covidvspneumonia/train\\\\pneumonia\\\\Viral Pneumonia (11).png',\n",
       " ...]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# verifying if the images have been extracted\n",
    "imagePaths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# assigining the labels to the images\n",
    "data = []\n",
    "labels = []\n",
    "# loop over the image paths\n",
    "for imagePath in imagePaths:\n",
    "    # extract the class label from the filename\n",
    "    label = imagePath.split(os.path.sep)[-2]\n",
    "    # load the image, swap color channels, and resize it to be a fixed\n",
    "    # 224x224 pixels while ignoring aspect ratio\n",
    "    image = cv2.imread(imagePath)\n",
    "    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
    "    image = cv2.resize(image, (64, 64))\n",
    "    # update the data and labels lists, respectively\n",
    "    data.append(image)\n",
    "    labels.append(label)\n",
    "# convert the data and labels to NumPy arrays while scaling the pixel\n",
    "# intensities to the range [0, 1]\n",
    "data = np.array(data) / 255.0\n",
    "labels = np.array(labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2213, 64, 64, 3)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# verifying the shape\n",
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['covid', 'covid', 'covid', ..., 'pneumonia', 'pneumonia',\n",
       "       'pneumonia'], dtype='<U9')"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# verifying the labels\n",
    "labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# perform one-hot encoding on the labels\n",
    "lb = LabelBinarizer()\n",
    "labels = lb.fit_transform(labels)\n",
    "labels = to_categorical(labels)\n",
    "# partition the data into training and testing splits using 80% of\n",
    "# the data for training and the remaining 20% for testing\n",
    "(x_train, x_test, y_train, y_test) = train_test_split(data, labels,\n",
    "    test_size=0.20, stratify=labels, random_state=42)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Augmentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Data Augmentation\n",
    "def process_data(x_train,y_train, x_test,y_test, batch_size):\n",
    "    #to prevent overfitting\n",
    "    train_datagen = ImageDataGenerator(shear_range=0.2,\n",
    "                                   rotation_range=20,\n",
    "                                   width_shift_range=0.2,\n",
    "                                   height_shift_range=0.2,\n",
    "                                   horizontal_flip=True,\n",
    "                                   vertical_flip=False,\n",
    "                                   zoom_range=0.2)\n",
    "\n",
    "\n",
    "    validation_datagen = ImageDataGenerator()\n",
    "\n",
    "    train_generator = train_datagen.flow(x_train,y_train,\n",
    "                                                        batch_size=batch_size)\n",
    "\n",
    "    validation_generator = validation_datagen.flow(x_test,y_test,\n",
    "                                                  batch_size=batch_size)\n",
    "\n",
    "\n",
    "\n",
    "    return train_generator, validation_generator "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Transfer Learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# build sequencial model\n",
    "def build_model(bottom_model, num_classes):\n",
    "    top_model = bottom_model.output\n",
    "    top_model = MaxPooling2D(pool_size=(2,2), strides=2)(top_model)\n",
    "    top_model = Flatten(name=\"flatten\")(top_model)\n",
    "    top_model = Dense(2, activation='sigmoid')(top_model)\n",
    "    model = Model(inputs=bottom_model.input, outputs=top_model)\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# adding top layer\n",
    "def addTopModel(bottom_model, num_classes):\n",
    "    top_model = bottom_model.output\n",
    "    top_model = MaxPooling2D(pool_size=(2,2), strides=2)(top_model)\n",
    "    top_model = Flatten(name=\"flatten\")(top_model)\n",
    "    top_model = Dense(512, activation=\"relu\")(top_model)\n",
    "    top_model = Dropout(0.5)(top_model)\n",
    "    top_model = Dense(2, activation='sigmoid')(top_model)\n",
    "    model = Model(inputs=bottom_model.input, outputs=top_model)\n",
    "    \n",
    "    return model\n",
    "\n",
    "\n",
    "def get_model(transferleaner,x_train,y_train, x_test,y_test):\n",
    "    model = addTopModel(m, num_classes) \n",
    "\n",
    "    for layer in m.layers:\n",
    "        layer.trainable = False\n",
    "\n",
    "\n",
    "        #data augmentation\n",
    "    train_generator, validation_generator = process_data(x_train,y_train,\n",
    "                                                     x_test,y_test, batch_size)\n",
    "\n",
    "\n",
    "    model.summary()\n",
    "\n",
    "    checkpoint = ModelCheckpoint(data_dir+'modelcnp' + '.h5',\n",
    "                                 monitor='val_loss',\n",
    "                                 mode=\"min\",\n",
    "                                 save_best_only=True,\n",
    "                                 verbose=1)\n",
    "\n",
    "    earlystop = EarlyStopping(monitor='val_loss',\n",
    "                              min_delta=0,\n",
    "                              patience=30,\n",
    "                              verbose=1,\n",
    "                              restore_best_weights=True)\n",
    "\n",
    "    learning_rate_reduction = ReduceLROnPlateau(monitor='val_loss',\n",
    "                                                patience=30,\n",
    "                                                verbose=1,\n",
    "                                                factor=0.8,\n",
    "                                                min_lr=0.0001,\n",
    "                                                mode=\"auto\",\n",
    "                                                min_delta=0.0001,\n",
    "                                                cooldown=5)\n",
    "\n",
    "    callbacks = [checkpoint, earlystop, learning_rate_reduction]\n",
    "\n",
    "    model.compile(loss=models_loss,\n",
    "                  optimizer=models_opt,\n",
    "                  metrics=['accuracy'])\n",
    "\n",
    "\n",
    "    history = model.fit(train_generator,\n",
    "                                    epochs=epochs,\n",
    "                                  callbacks=callbacks,\n",
    "                                  validation_data=(x_test, y_test),\n",
    "                                  validation_steps=len(x_test) / batch_size,\n",
    "                                  steps_per_epoch=len(x_train) / batch_size)\n",
    "    return model, history"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plotting confusion matrix for testing\n",
    "def plot_confusion_matrix(model, x_test, y_test):\n",
    "    fig, ax = plt.subplots(figsize=(8,6))\n",
    "\n",
    "    classes =  ['PNEUMONIA','COVID']\n",
    "    \n",
    "    y_pred = model.predict(x_test, batch_size=batch_size)\n",
    "    y_pred = np.argmax(y_pred, axis=1)\n",
    "    y_test = np.argmax(y_test, axis=1) \n",
    "    \n",
    "    print('Confusion Matrix')\n",
    "    cm = confusion_matrix(y_test, y_pred, normalize='true')\n",
    "    #cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    " \n",
    "    sns.heatmap(cm,cmap='Purples',fmt='g', annot=True)\n",
    "    tick_marks = [0.5,1.5]\n",
    "    plt.xticks(tick_marks, classes)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    plt.title('Confusion Matrix - Normalized')\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "    bottom, top = ax.get_ylim()\n",
    "    \n",
    "    ax.set_ylim(bottom + 0.5, top - 0.5)\n",
    "    \n",
    "    plt.show()\n",
    "    \n",
    "    acc = accuracy_score(y_test, y_pred)\n",
    "    precision = precision_score(y_test, y_pred, average=None)\n",
    "    recall = recall_score(y_test, y_pred, average=None)\n",
    "    f1 = f1_score(y_test, y_pred, average=None)\n",
    "    \n",
    "    print(\"Precision Score: {}\".format(precision))\n",
    "    print(\"Recall Score: {}\".format(recall))\n",
    "    print(\"F1 Score: {}\".format(f1))\n",
    "    print(\"Accuracy Score: {}\".format(acc))\n",
    "\n",
    "    return plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ROC AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#ROC curve\n",
    "def multiclass_roc_auc_score(x_test, y_test, model, average=\"micro\"):\n",
    "    \n",
    "\n",
    "    y_pred = model.predict(x_test, batch_size=batch_size)\n",
    "    \n",
    "    # Convert to Binary classes\n",
    "    y_pred_bin = np.argmax(y_pred, axis=1)\n",
    "    y_test_bin = np.argmax(y_test, axis=1)\n",
    "\n",
    "    \n",
    "    fpr, tpr, thresholds = roc_curve(y_test_bin, y_pred_bin)\n",
    "    \n",
    "    auc_keras = auc(fpr, tpr)\n",
    "    \n",
    "    print('AUC: {}'.format(auc_keras))    \n",
    "    print('Log Loss: {}'.format(metrics.log_loss(y_test.argmax(axis=1), y_pred)))\n",
    "\n",
    "        \n",
    "    plt.plot(fpr, tpr,  color='darkorange', label='ROC curve (area = %0.2f)' % auc_keras)\n",
    "    plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')\n",
    "    plt.xlim([0.0, 1.0])\n",
    "    plt.ylim([0.0, 1.0])\n",
    "    plt.rcParams['font.size'] = 12\n",
    "    plt.title('ROC curve for our model')\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    plt.legend(loc='best', fancybox=True)\n",
    "    plt.grid(True)\n",
    "\n",
    "\n",
    "    return plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model loss\n",
    "def plot_learning_curves(r):\n",
    "    plt.figure(figsize=(12,4))\n",
    "    \n",
    "    plt.subplot(1,2,1)\n",
    "    plt.plot(r.history['loss'])\n",
    "    plt.plot(r.history['val_loss'])\n",
    "    plt.title('model loss')\n",
    "    plt.ylabel('loss')\n",
    "    plt.xlabel('epoch')\n",
    "    plt.grid(True)\n",
    "    \n",
    "    plt.subplot(1,2,2)\n",
    "    plt.plot(r.history['accuracy'])\n",
    "    plt.plot(r.history['val_accuracy'])\n",
    "    plt.title('model accuracy')\n",
    "    plt.ylabel('accuracy')\n",
    "    plt.xlabel('epoch')\n",
    "    plt.legend(['train', 'test'], loc='upper left')\n",
    "    plt.grid(True)\n",
    "    plt.tight_layout()\n",
    "    \n",
    "    return plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Assigning Parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Moo8-9Hedbty"
   },
   "outputs": [],
   "source": [
    "#parameters\n",
    "data_dir = '/'\n",
    "version = '-v5-'\n",
    "num_classes = 2\n",
    "ig_size= 64\n",
    "epochs = 100\n",
    "batch_size = 32\n",
    "\n",
    "models_loss = 'binary_crossentropy'\n",
    "models_opt = 'adam' #Adam(lr=0.001, decay=0.001/600) #SGD(learning_rate=0.001, momentum=0.9) #ADAM(lr=0.001)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Modeling - Base CNN Model\n",
    "\n",
    "First, I used the sequential keras model which is th easiest way to build a model in keras. It allows one to build a model, layer by layer. My first layer will be a Conv2D layer that will deal with my input x-ray images of shape 64, 64, 3 using 64 nodes. The 3 in the shape sigifying that the images are in RGB. The kernel size is set to 3, which means the filter that would be used to scan through every pixel will be a matrix of 3x3. \n",
    "\n",
    "Next, a MaxPooling2D layer was utilized in order to reduce overfitting and dimentionality i.e in order to reduce the number of parameters to learn from and the amount of comuputation performed in the network. I set it to divide the dimentions by a matrix of 2x2 while moving 2 strides after choosing the maximum value of each patch. After this, I flattened the output from the MaxPooling layer in order to make it linear and pass it to the Dense layer implemented, which uses a node of 2 which represents the number of output classes to be predicted."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_input (InputLayer)    [(None, 64, 64, 3)]       0         \n",
      "_________________________________________________________________\n",
      "conv2d (Conv2D)              (None, 62, 62, 64)        1792      \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 31, 31, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 61504)             0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 2)                 123010    \n",
      "=================================================================\n",
      "Total params: 124,802\n",
      "Trainable params: 123,010\n",
      "Non-trainable params: 1,792\n",
      "_________________________________________________________________\n",
      "Epoch 1/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.4213 - accuracy: 0.8226\n",
      "Epoch 00001: val_loss improved from inf to 0.21483, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 6s 107ms/step - loss: 0.4213 - accuracy: 0.8226 - val_loss: 0.2148 - val_accuracy: 0.9300 - lr: 0.0010\n",
      "Epoch 2/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3223 - accuracy: 0.8746\n",
      "Epoch 00002: val_loss improved from 0.21483 to 0.19551, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 97ms/step - loss: 0.3223 - accuracy: 0.8746 - val_loss: 0.1955 - val_accuracy: 0.9345 - lr: 0.0010\n",
      "Epoch 3/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2967 - accuracy: 0.8774\n",
      "Epoch 00003: val_loss improved from 0.19551 to 0.17436, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 98ms/step - loss: 0.2967 - accuracy: 0.8774 - val_loss: 0.1744 - val_accuracy: 0.9345 - lr: 0.0010\n",
      "Epoch 4/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2975 - accuracy: 0.8842\n",
      "Epoch 00004: val_loss improved from 0.17436 to 0.15852, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 6s 98ms/step - loss: 0.2975 - accuracy: 0.8842 - val_loss: 0.1585 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 5/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2765 - accuracy: 0.8847\n",
      "Epoch 00005: val_loss did not improve from 0.15852\n",
      "56/55 [==============================] - 5s 98ms/step - loss: 0.2765 - accuracy: 0.8847 - val_loss: 0.1713 - val_accuracy: 0.9481 - lr: 0.0010\n",
      "Epoch 6/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3083 - accuracy: 0.8814\n",
      "Epoch 00006: val_loss did not improve from 0.15852\n",
      "56/55 [==============================] - 5s 97ms/step - loss: 0.3083 - accuracy: 0.8814 - val_loss: 0.1633 - val_accuracy: 0.9481 - lr: 0.0010\n",
      "Epoch 7/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2707 - accuracy: 0.8915\n",
      "Epoch 00007: val_loss improved from 0.15852 to 0.14697, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 6s 107ms/step - loss: 0.2707 - accuracy: 0.8915 - val_loss: 0.1470 - val_accuracy: 0.9526 - lr: 0.0010\n",
      "Epoch 8/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2616 - accuracy: 0.8955\n",
      "Epoch 00008: val_loss improved from 0.14697 to 0.14107, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 98ms/step - loss: 0.2616 - accuracy: 0.8955 - val_loss: 0.1411 - val_accuracy: 0.9458 - lr: 0.0010\n",
      "Epoch 9/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2432 - accuracy: 0.9045\n",
      "Epoch 00009: val_loss improved from 0.14107 to 0.13235, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 6s 100ms/step - loss: 0.2432 - accuracy: 0.9045 - val_loss: 0.1324 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 10/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2517 - accuracy: 0.9011\n",
      "Epoch 00010: val_loss did not improve from 0.13235\n",
      "56/55 [==============================] - 5s 95ms/step - loss: 0.2517 - accuracy: 0.9011 - val_loss: 0.1752 - val_accuracy: 0.9436 - lr: 0.0010\n",
      "Epoch 11/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2717 - accuracy: 0.8915\n",
      "Epoch 00011: val_loss did not improve from 0.13235\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.2717 - accuracy: 0.8915 - val_loss: 0.1359 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 12/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2864 - accuracy: 0.8847\n",
      "Epoch 00012: val_loss did not improve from 0.13235\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2864 - accuracy: 0.8847 - val_loss: 0.1545 - val_accuracy: 0.9458 - lr: 0.0010\n",
      "Epoch 13/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2557 - accuracy: 0.9006\n",
      "Epoch 00013: val_loss improved from 0.13235 to 0.12959, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.2557 - accuracy: 0.9006 - val_loss: 0.1296 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 14/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2463 - accuracy: 0.9023\n",
      "Epoch 00014: val_loss improved from 0.12959 to 0.12632, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 92ms/step - loss: 0.2463 - accuracy: 0.9023 - val_loss: 0.1263 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 15/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2634 - accuracy: 0.8955\n",
      "Epoch 00015: val_loss did not improve from 0.12632\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2634 - accuracy: 0.8955 - val_loss: 0.1289 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 16/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2773 - accuracy: 0.8910\n",
      "Epoch 00016: val_loss did not improve from 0.12632\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2773 - accuracy: 0.8910 - val_loss: 0.1574 - val_accuracy: 0.9391 - lr: 0.0010\n",
      "Epoch 17/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2660 - accuracy: 0.9028\n",
      "Epoch 00017: val_loss did not improve from 0.12632\n",
      "56/55 [==============================] - 5s 91ms/step - loss: 0.2660 - accuracy: 0.9028 - val_loss: 0.1425 - val_accuracy: 0.9526 - lr: 0.0010\n",
      "Epoch 18/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2308 - accuracy: 0.9113\n",
      "Epoch 00018: val_loss improved from 0.12632 to 0.12248, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 86ms/step - loss: 0.2308 - accuracy: 0.9113 - val_loss: 0.1225 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 19/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2555 - accuracy: 0.9017\n",
      "Epoch 00019: val_loss did not improve from 0.12248\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2555 - accuracy: 0.9017 - val_loss: 0.1227 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 20/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3085 - accuracy: 0.8763\n",
      "Epoch 00020: val_loss did not improve from 0.12248\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.3085 - accuracy: 0.8763 - val_loss: 0.1434 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 21/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2516 - accuracy: 0.9011\n",
      "Epoch 00021: val_loss did not improve from 0.12248\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2516 - accuracy: 0.9011 - val_loss: 0.1253 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 22/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2458 - accuracy: 0.9068\n",
      "Epoch 00022: val_loss did not improve from 0.12248\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2458 - accuracy: 0.9068 - val_loss: 0.1266 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 23/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2538 - accuracy: 0.9107\n",
      "Epoch 00023: val_loss did not improve from 0.12248\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2538 - accuracy: 0.9107 - val_loss: 0.1462 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 24/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2532 - accuracy: 0.9000\n",
      "Epoch 00024: val_loss did not improve from 0.12248\n",
      "56/55 [==============================] - 5s 87ms/step - loss: 0.2532 - accuracy: 0.9000 - val_loss: 0.1229 - val_accuracy: 0.9571 - lr: 0.0010\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 25/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2689 - accuracy: 0.8910\n",
      "Epoch 00025: val_loss improved from 0.12248 to 0.12231, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 84ms/step - loss: 0.2689 - accuracy: 0.8910 - val_loss: 0.1223 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 26/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2639 - accuracy: 0.8983\n",
      "Epoch 00026: val_loss did not improve from 0.12231\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2639 - accuracy: 0.8983 - val_loss: 0.1284 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 27/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2451 - accuracy: 0.9062\n",
      "Epoch 00027: val_loss improved from 0.12231 to 0.11988, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 86ms/step - loss: 0.2451 - accuracy: 0.9062 - val_loss: 0.1199 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 28/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2395 - accuracy: 0.9034\n",
      "Epoch 00028: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2395 - accuracy: 0.9034 - val_loss: 0.1818 - val_accuracy: 0.9278 - lr: 0.0010\n",
      "Epoch 29/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3155 - accuracy: 0.8734\n",
      "Epoch 00029: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.3155 - accuracy: 0.8734 - val_loss: 0.1496 - val_accuracy: 0.9458 - lr: 0.0010\n",
      "Epoch 30/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2544 - accuracy: 0.9040\n",
      "Epoch 00030: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2544 - accuracy: 0.9040 - val_loss: 0.1253 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 31/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2708 - accuracy: 0.8927\n",
      "Epoch 00031: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2708 - accuracy: 0.8927 - val_loss: 0.1383 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 32/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2697 - accuracy: 0.8904\n",
      "Epoch 00032: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2697 - accuracy: 0.8904 - val_loss: 0.1257 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 33/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2338 - accuracy: 0.9045\n",
      "Epoch 00033: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2338 - accuracy: 0.9045 - val_loss: 0.1397 - val_accuracy: 0.9526 - lr: 0.0010\n",
      "Epoch 34/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2174 - accuracy: 0.9181\n",
      "Epoch 00034: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2174 - accuracy: 0.9181 - val_loss: 0.1279 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 35/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2310 - accuracy: 0.9147\n",
      "Epoch 00035: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 84ms/step - loss: 0.2310 - accuracy: 0.9147 - val_loss: 0.1564 - val_accuracy: 0.9413 - lr: 0.0010\n",
      "Epoch 36/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3249 - accuracy: 0.8723\n",
      "Epoch 00036: val_loss did not improve from 0.11988\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.3249 - accuracy: 0.8723 - val_loss: 0.1212 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 37/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2372 - accuracy: 0.9068\n",
      "Epoch 00037: val_loss improved from 0.11988 to 0.11378, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 87ms/step - loss: 0.2372 - accuracy: 0.9068 - val_loss: 0.1138 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 38/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2477 - accuracy: 0.9045\n",
      "Epoch 00038: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 85ms/step - loss: 0.2477 - accuracy: 0.9045 - val_loss: 0.1382 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 39/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2562 - accuracy: 0.8966\n",
      "Epoch 00039: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2562 - accuracy: 0.8966 - val_loss: 0.1318 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 40/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2217 - accuracy: 0.9119\n",
      "Epoch 00040: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2217 - accuracy: 0.9119 - val_loss: 0.1297 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 41/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2443 - accuracy: 0.9006\n",
      "Epoch 00041: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2443 - accuracy: 0.9006 - val_loss: 0.1618 - val_accuracy: 0.9413 - lr: 0.0010\n",
      "Epoch 42/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2689 - accuracy: 0.8966\n",
      "Epoch 00042: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2689 - accuracy: 0.8966 - val_loss: 0.1343 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 43/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2381 - accuracy: 0.9096\n",
      "Epoch 00043: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2381 - accuracy: 0.9096 - val_loss: 0.1204 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 44/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2266 - accuracy: 0.9209\n",
      "Epoch 00044: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2266 - accuracy: 0.9209 - val_loss: 0.1295 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 45/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2218 - accuracy: 0.9085\n",
      "Epoch 00045: val_loss did not improve from 0.11378\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2218 - accuracy: 0.9085 - val_loss: 0.1203 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 46/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2196 - accuracy: 0.9056\n",
      "Epoch 00046: val_loss improved from 0.11378 to 0.11238, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 85ms/step - loss: 0.2196 - accuracy: 0.9056 - val_loss: 0.1124 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 47/100\n",
      "55/55 [============================>.] - ETA: 0s - loss: 0.2410 - accuracy: 0.9074\n",
      "Epoch 00047: val_loss did not improve from 0.11238\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2409 - accuracy: 0.9073 - val_loss: 0.1785 - val_accuracy: 0.9300 - lr: 0.0010\n",
      "Epoch 48/100\n",
      "55/55 [============================>.] - ETA: 0s - loss: 0.2559 - accuracy: 0.8983\n",
      "Epoch 00048: val_loss did not improve from 0.11238\n",
      "56/55 [==============================] - 5s 84ms/step - loss: 0.2553 - accuracy: 0.8989 - val_loss: 0.1532 - val_accuracy: 0.9391 - lr: 0.0010\n",
      "Epoch 49/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2251 - accuracy: 0.9130\n",
      "Epoch 00049: val_loss improved from 0.11238 to 0.11147, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 87ms/step - loss: 0.2251 - accuracy: 0.9130 - val_loss: 0.1115 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 50/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2647 - accuracy: 0.8881\n",
      "Epoch 00050: val_loss did not improve from 0.11147\n",
      "56/55 [==============================] - 5s 82ms/step - loss: 0.2647 - accuracy: 0.8881 - val_loss: 0.1760 - val_accuracy: 0.9345 - lr: 0.0010\n",
      "Epoch 51/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2841 - accuracy: 0.8893\n",
      "Epoch 00051: val_loss did not improve from 0.11147\n",
      "56/55 [==============================] - 5s 86ms/step - loss: 0.2841 - accuracy: 0.8893 - val_loss: 0.1259 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 52/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2633 - accuracy: 0.8989\n",
      "Epoch 00052: val_loss did not improve from 0.11147\n",
      "56/55 [==============================] - 5s 90ms/step - loss: 0.2633 - accuracy: 0.8989 - val_loss: 0.1771 - val_accuracy: 0.9368 - lr: 0.0010\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 53/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2508 - accuracy: 0.8960\n",
      "Epoch 00053: val_loss improved from 0.11147 to 0.10751, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 97ms/step - loss: 0.2508 - accuracy: 0.8960 - val_loss: 0.1075 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 54/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2364 - accuracy: 0.9051\n",
      "Epoch 00054: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2364 - accuracy: 0.9051 - val_loss: 0.1398 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 55/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2405 - accuracy: 0.9000\n",
      "Epoch 00055: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 84ms/step - loss: 0.2405 - accuracy: 0.9000 - val_loss: 0.1089 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 56/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2313 - accuracy: 0.9073\n",
      "Epoch 00056: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 90ms/step - loss: 0.2313 - accuracy: 0.9073 - val_loss: 0.1143 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 57/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2533 - accuracy: 0.9023\n",
      "Epoch 00057: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2533 - accuracy: 0.9023 - val_loss: 0.1088 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 58/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3119 - accuracy: 0.8701\n",
      "Epoch 00058: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.3119 - accuracy: 0.8701 - val_loss: 0.1304 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 59/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2385 - accuracy: 0.9102 ETA: 0s - loss: 0.2418 - accuracy\n",
      "Epoch 00059: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 92ms/step - loss: 0.2385 - accuracy: 0.9102 - val_loss: 0.1228 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 60/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2157 - accuracy: 0.9192\n",
      "Epoch 00060: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2157 - accuracy: 0.9192 - val_loss: 0.1079 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 61/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2128 - accuracy: 0.9141\n",
      "Epoch 00061: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 86ms/step - loss: 0.2128 - accuracy: 0.9141 - val_loss: 0.1361 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 62/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2219 - accuracy: 0.9107\n",
      "Epoch 00062: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 90ms/step - loss: 0.2219 - accuracy: 0.9107 - val_loss: 0.1128 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 63/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2208 - accuracy: 0.9079\n",
      "Epoch 00063: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 88ms/step - loss: 0.2208 - accuracy: 0.9079 - val_loss: 0.1270 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 64/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2123 - accuracy: 0.9186\n",
      "Epoch 00064: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 86ms/step - loss: 0.2123 - accuracy: 0.9186 - val_loss: 0.1639 - val_accuracy: 0.9368 - lr: 0.0010\n",
      "Epoch 65/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2039 - accuracy: 0.9203\n",
      "Epoch 00065: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 92ms/step - loss: 0.2039 - accuracy: 0.9203 - val_loss: 0.1175 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 66/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2269 - accuracy: 0.9073\n",
      "Epoch 00066: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 95ms/step - loss: 0.2269 - accuracy: 0.9073 - val_loss: 0.1146 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 67/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2074 - accuracy: 0.9220\n",
      "Epoch 00067: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 97ms/step - loss: 0.2074 - accuracy: 0.9220 - val_loss: 0.1135 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 68/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2043 - accuracy: 0.9203\n",
      "Epoch 00068: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2043 - accuracy: 0.9203 - val_loss: 0.1134 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 69/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1989 - accuracy: 0.9181\n",
      "Epoch 00069: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 6s 102ms/step - loss: 0.1989 - accuracy: 0.9181 - val_loss: 0.1335 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 70/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2358 - accuracy: 0.9079\n",
      "Epoch 00070: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 97ms/step - loss: 0.2358 - accuracy: 0.9079 - val_loss: 0.1747 - val_accuracy: 0.9323 - lr: 0.0010\n",
      "Epoch 71/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2163 - accuracy: 0.9175\n",
      "Epoch 00071: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 95ms/step - loss: 0.2163 - accuracy: 0.9175 - val_loss: 0.1235 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 72/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2322 - accuracy: 0.9096\n",
      "Epoch 00072: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 6s 99ms/step - loss: 0.2322 - accuracy: 0.9096 - val_loss: 0.1887 - val_accuracy: 0.9255 - lr: 0.0010\n",
      "Epoch 73/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2387 - accuracy: 0.9079\n",
      "Epoch 00073: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.2387 - accuracy: 0.9079 - val_loss: 0.2116 - val_accuracy: 0.9187 - lr: 0.0010\n",
      "Epoch 74/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2383 - accuracy: 0.9107\n",
      "Epoch 00074: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2383 - accuracy: 0.9107 - val_loss: 0.1163 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 75/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2592 - accuracy: 0.8898\n",
      "Epoch 00075: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 88ms/step - loss: 0.2592 - accuracy: 0.8898 - val_loss: 0.1165 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 76/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2116 - accuracy: 0.9164\n",
      "Epoch 00076: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 95ms/step - loss: 0.2116 - accuracy: 0.9164 - val_loss: 0.1256 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 77/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2354 - accuracy: 0.9096\n",
      "Epoch 00077: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 90ms/step - loss: 0.2354 - accuracy: 0.9096 - val_loss: 0.1650 - val_accuracy: 0.9300 - lr: 0.0010\n",
      "Epoch 78/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2226 - accuracy: 0.9079\n",
      "Epoch 00078: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 92ms/step - loss: 0.2226 - accuracy: 0.9079 - val_loss: 0.1484 - val_accuracy: 0.9436 - lr: 0.0010\n",
      "Epoch 79/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2206 - accuracy: 0.9119\n",
      "Epoch 00079: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 5s 96ms/step - loss: 0.2206 - accuracy: 0.9119 - val_loss: 0.1438 - val_accuracy: 0.9503 - lr: 0.0010\n",
      "Epoch 80/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2282 - accuracy: 0.9119\n",
      "Epoch 00080: val_loss did not improve from 0.10751\n",
      "56/55 [==============================] - 6s 101ms/step - loss: 0.2282 - accuracy: 0.9119 - val_loss: 0.1110 - val_accuracy: 0.9684 - lr: 0.0010\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 81/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2271 - accuracy: 0.9096\n",
      "Epoch 00081: val_loss improved from 0.10751 to 0.10735, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 95ms/step - loss: 0.2271 - accuracy: 0.9096 - val_loss: 0.1074 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 82/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2499 - accuracy: 0.9011\n",
      "Epoch 00082: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 92ms/step - loss: 0.2499 - accuracy: 0.9011 - val_loss: 0.1198 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 83/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2024 - accuracy: 0.9226\n",
      "Epoch 00083: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2024 - accuracy: 0.9226 - val_loss: 0.1098 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 84/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1884 - accuracy: 0.9271\n",
      "Epoch 00084: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.1884 - accuracy: 0.9271 - val_loss: 0.1076 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 85/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2032 - accuracy: 0.9282\n",
      "Epoch 00085: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2032 - accuracy: 0.9282 - val_loss: 0.1362 - val_accuracy: 0.9526 - lr: 0.0010\n",
      "Epoch 86/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2437 - accuracy: 0.9011\n",
      "Epoch 00086: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.2437 - accuracy: 0.9011 - val_loss: 0.1112 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 87/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2143 - accuracy: 0.9192\n",
      "Epoch 00087: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2143 - accuracy: 0.9192 - val_loss: 0.1126 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 88/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2616 - accuracy: 0.8910\n",
      "Epoch 00088: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 87ms/step - loss: 0.2616 - accuracy: 0.8910 - val_loss: 0.1199 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 89/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2311 - accuracy: 0.9102\n",
      "Epoch 00089: val_loss did not improve from 0.10735\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.2311 - accuracy: 0.9102 - val_loss: 0.1115 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 90/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2072 - accuracy: 0.9169\n",
      "Epoch 00090: val_loss improved from 0.10735 to 0.10594, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 6s 100ms/step - loss: 0.2072 - accuracy: 0.9169 - val_loss: 0.1059 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 91/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2334 - accuracy: 0.9034\n",
      "Epoch 00091: val_loss improved from 0.10594 to 0.10589, saving model to /sequentialpnd.h5\n",
      "56/55 [==============================] - 5s 93ms/step - loss: 0.2334 - accuracy: 0.9034 - val_loss: 0.1059 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 92/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2173 - accuracy: 0.9102\n",
      "Epoch 00092: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 92ms/step - loss: 0.2173 - accuracy: 0.9102 - val_loss: 0.1125 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 93/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2390 - accuracy: 0.9136\n",
      "Epoch 00093: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2390 - accuracy: 0.9136 - val_loss: 0.1076 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 94/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2078 - accuracy: 0.9186\n",
      "Epoch 00094: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 86ms/step - loss: 0.2078 - accuracy: 0.9186 - val_loss: 0.1124 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 95/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2029 - accuracy: 0.9215\n",
      "Epoch 00095: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 90ms/step - loss: 0.2029 - accuracy: 0.9215 - val_loss: 0.2139 - val_accuracy: 0.9120 - lr: 0.0010\n",
      "Epoch 96/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1976 - accuracy: 0.9249\n",
      "Epoch 00096: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 88ms/step - loss: 0.1976 - accuracy: 0.9249 - val_loss: 0.1175 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 97/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2001 - accuracy: 0.9254\n",
      "Epoch 00097: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 89ms/step - loss: 0.2001 - accuracy: 0.9254 - val_loss: 0.1104 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 98/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2117 - accuracy: 0.9164\n",
      "Epoch 00098: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 94ms/step - loss: 0.2117 - accuracy: 0.9164 - val_loss: 0.1180 - val_accuracy: 0.9549 - lr: 0.0010\n",
      "Epoch 99/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2354 - accuracy: 0.8994\n",
      "Epoch 00099: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 84ms/step - loss: 0.2354 - accuracy: 0.8994 - val_loss: 0.1085 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 100/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2254 - accuracy: 0.9153\n",
      "Epoch 00100: val_loss did not improve from 0.10589\n",
      "56/55 [==============================] - 5s 83ms/step - loss: 0.2254 - accuracy: 0.9153 - val_loss: 0.1135 - val_accuracy: 0.9684 - lr: 0.0010\n"
     ]
    }
   ],
   "source": [
    "# calling thesequential model\n",
    "m = Sequential(Conv2D(ig_size, kernel_size=3, activation='relu', \n",
    "                      input_shape=(64,64,3)))\n",
    "\n",
    "model = build_model(m, num_classes) \n",
    "\n",
    "for layer in m.layers:\n",
    "    layer.trainable = False\n",
    " \n",
    "\n",
    " #data augmentation\n",
    "train_generator, validation_generator = process_data(x_train,y_train,\n",
    "                                                 x_test,y_test, batch_size)\n",
    "model.summary()\n",
    "\n",
    "checkpoint = ModelCheckpoint(data_dir+'sequentialpnd' + '.h5',\n",
    "                             monitor='val_loss',\n",
    "                             mode=\"min\",\n",
    "                             save_best_only=True,\n",
    "                             verbose=1)\n",
    "\n",
    "earlystop = EarlyStopping(monitor='val_loss',\n",
    "                          min_delta=0,\n",
    "                          patience=30,\n",
    "                          verbose=1,\n",
    "                          restore_best_weights=True)\n",
    "\n",
    "learning_rate_reduction = ReduceLROnPlateau(monitor='val_loss',\n",
    "                                            patience=30,\n",
    "                                            verbose=1,\n",
    "                                            factor=0.8,\n",
    "                                            min_lr=0.0001,\n",
    "                                            mode=\"auto\",\n",
    "                                            min_delta=0.0001,\n",
    "                                            cooldown=5)\n",
    "\n",
    "\n",
    "callbacks = [checkpoint, earlystop, learning_rate_reduction]\n",
    "\n",
    "model.compile(loss=models_loss,\n",
    "              optimizer=models_opt,\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "\n",
    "history = model.fit(train_generator,\n",
    "                             epochs=epochs,\n",
    "                               callbacks=callbacks,\n",
    "                               validation_data=validation_generator,\n",
    "                              validation_steps=len(x_test) / batch_size,\n",
    "                              steps_per_epoch=len(x_train) / batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: [0.94972067 0.98106061]\n",
      "Recall Score: [0.97142857 0.96641791]\n",
      "F1 Score: [0.96045198 0.97368421]\n",
      "Accuracy Score: 0.9683972911963883\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9689232409381663\n",
      "Log Loss: 0.11345926234635331\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Confusion matrix\n",
    "plot_confusion_matrix(model, x_test, y_test)\n",
    "## Learning curve\n",
    "plot_learning_curves(history)\n",
    "## ROC AUC\n",
    "multiclass_roc_auc_score(x_test, y_test, model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# saving model\n",
    "model.save(\"sequential.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "model = load_model('./sequential.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Modeling - Using VGG19\n",
    "\n",
    "VGG-19 is a transfer learning alorithm which means its an algorithm with 16 convolutional layers that focuses on storing knowledge that can be applied to different but relaed problems."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 64, 64, 3)]       0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 64, 64, 64)        1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 64, 64, 64)        36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 32, 32, 64)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 32, 32, 128)       73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 32, 32, 128)       147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 16, 16, 128)       0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 16, 16, 256)       295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 16, 16, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 16, 16, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_conv4 (Conv2D)        (None, 16, 16, 256)       590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 8, 8, 256)         0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 8, 8, 512)         1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 8, 8, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 8, 8, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv4 (Conv2D)        (None, 8, 8, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 4, 4, 512)         0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 4, 4, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 4, 4, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 4, 4, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv4 (Conv2D)        (None, 4, 4, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 2, 2, 512)         0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 1, 1, 512)         0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 2)                 1026      \n",
      "=================================================================\n",
      "Total params: 20,288,066\n",
      "Trainable params: 263,682\n",
      "Non-trainable params: 20,024,384\n",
      "_________________________________________________________________\n",
      "Epoch 1/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.3981 - accuracy: 0.8215\n",
      "Epoch 00001: val_loss improved from inf to 0.17578, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.3981 - accuracy: 0.8215 - val_loss: 0.1758 - val_accuracy: 0.9413 - lr: 0.0010\n",
      "Epoch 2/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2414 - accuracy: 0.9136\n",
      "Epoch 00002: val_loss improved from 0.17578 to 0.12758, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 70s 1s/step - loss: 0.2414 - accuracy: 0.9136 - val_loss: 0.1276 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 3/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2182 - accuracy: 0.9158\n",
      "Epoch 00003: val_loss improved from 0.12758 to 0.11540, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 70s 1s/step - loss: 0.2182 - accuracy: 0.9158 - val_loss: 0.1154 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 4/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2025 - accuracy: 0.9215\n",
      "Epoch 00004: val_loss did not improve from 0.11540\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.2025 - accuracy: 0.9215 - val_loss: 0.1182 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 5/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1962 - accuracy: 0.9232\n",
      "Epoch 00005: val_loss did not improve from 0.11540\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1962 - accuracy: 0.9232 - val_loss: 0.1208 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 6/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2024 - accuracy: 0.9232\n",
      "Epoch 00006: val_loss did not improve from 0.11540\n",
      "56/55 [==============================] - 71s 1s/step - loss: 0.2024 - accuracy: 0.9232 - val_loss: 0.1274 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 7/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1939 - accuracy: 0.9237\n",
      "Epoch 00007: val_loss improved from 0.11540 to 0.09602, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1939 - accuracy: 0.9237 - val_loss: 0.0960 - val_accuracy: 0.9707 - lr: 0.0010\n",
      "Epoch 8/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1978 - accuracy: 0.9203\n",
      "Epoch 00008: val_loss did not improve from 0.09602\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1978 - accuracy: 0.9203 - val_loss: 0.1051 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 9/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1942 - accuracy: 0.9299\n",
      "Epoch 00009: val_loss improved from 0.09602 to 0.09027, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 70s 1s/step - loss: 0.1942 - accuracy: 0.9299 - val_loss: 0.0903 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 10/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1996 - accuracy: 0.9175\n",
      "Epoch 00010: val_loss improved from 0.09027 to 0.08720, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 70s 1s/step - loss: 0.1996 - accuracy: 0.9175 - val_loss: 0.0872 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 11/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.2035 - accuracy: 0.9141\n",
      "Epoch 00011: val_loss did not improve from 0.08720\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.2035 - accuracy: 0.9141 - val_loss: 0.1459 - val_accuracy: 0.9526 - lr: 0.0010\n",
      "Epoch 12/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1767 - accuracy: 0.9395\n",
      "Epoch 00012: val_loss did not improve from 0.08720\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1767 - accuracy: 0.9395 - val_loss: 0.0891 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 13/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1683 - accuracy: 0.9418\n",
      "Epoch 00013: val_loss did not improve from 0.08720\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1683 - accuracy: 0.9418 - val_loss: 0.1019 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 14/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1890 - accuracy: 0.9243\n",
      "Epoch 00014: val_loss did not improve from 0.08720\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1890 - accuracy: 0.9243 - val_loss: 0.1097 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 15/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "56/55 [==============================] - ETA: 0s - loss: 0.1745 - accuracy: 0.9237\n",
      "Epoch 00015: val_loss did not improve from 0.08720\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1745 - accuracy: 0.9237 - val_loss: 0.0979 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 16/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1664 - accuracy: 0.9384\n",
      "Epoch 00016: val_loss did not improve from 0.08720\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1664 - accuracy: 0.9384 - val_loss: 0.0974 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 17/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1445 - accuracy: 0.9475\n",
      "Epoch 00017: val_loss improved from 0.08720 to 0.08581, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1445 - accuracy: 0.9475 - val_loss: 0.0858 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 18/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1772 - accuracy: 0.9254\n",
      "Epoch 00018: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1772 - accuracy: 0.9254 - val_loss: 0.1309 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 19/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1729 - accuracy: 0.9339\n",
      "Epoch 00019: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1729 - accuracy: 0.9339 - val_loss: 0.1277 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 20/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1691 - accuracy: 0.9322\n",
      "Epoch 00020: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1691 - accuracy: 0.9322 - val_loss: 0.0952 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 21/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1696 - accuracy: 0.9395\n",
      "Epoch 00021: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1696 - accuracy: 0.9395 - val_loss: 0.1072 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 22/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1655 - accuracy: 0.9339\n",
      "Epoch 00022: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1655 - accuracy: 0.9339 - val_loss: 0.1142 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 23/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1667 - accuracy: 0.9362\n",
      "Epoch 00023: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1667 - accuracy: 0.9362 - val_loss: 0.0936 - val_accuracy: 0.9707 - lr: 0.0010\n",
      "Epoch 24/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1778 - accuracy: 0.9311\n",
      "Epoch 00024: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1778 - accuracy: 0.9311 - val_loss: 0.1256 - val_accuracy: 0.9571 - lr: 0.0010\n",
      "Epoch 25/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1742 - accuracy: 0.9282\n",
      "Epoch 00025: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1742 - accuracy: 0.9282 - val_loss: 0.1049 - val_accuracy: 0.9707 - lr: 0.0010\n",
      "Epoch 26/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1625 - accuracy: 0.9395\n",
      "Epoch 00026: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1625 - accuracy: 0.9395 - val_loss: 0.0948 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 27/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1688 - accuracy: 0.9299\n",
      "Epoch 00027: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1688 - accuracy: 0.9299 - val_loss: 0.1037 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 28/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1639 - accuracy: 0.9339\n",
      "Epoch 00028: val_loss did not improve from 0.08581\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1639 - accuracy: 0.9339 - val_loss: 0.1004 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 29/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1760 - accuracy: 0.9316\n",
      "Epoch 00029: val_loss improved from 0.08581 to 0.08295, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1760 - accuracy: 0.9316 - val_loss: 0.0829 - val_accuracy: 0.9774 - lr: 0.0010\n",
      "Epoch 30/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1523 - accuracy: 0.9362\n",
      "Epoch 00030: val_loss improved from 0.08295 to 0.07471, saving model to /modelcnp.h5\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1523 - accuracy: 0.9362 - val_loss: 0.0747 - val_accuracy: 0.9774 - lr: 0.0010\n",
      "Epoch 31/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1702 - accuracy: 0.9322\n",
      "Epoch 00031: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1702 - accuracy: 0.9322 - val_loss: 0.0960 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 32/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1619 - accuracy: 0.9350\n",
      "Epoch 00032: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1619 - accuracy: 0.9350 - val_loss: 0.1059 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 33/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1723 - accuracy: 0.9339\n",
      "Epoch 00033: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1723 - accuracy: 0.9339 - val_loss: 0.0971 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 34/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1429 - accuracy: 0.9458\n",
      "Epoch 00034: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1429 - accuracy: 0.9458 - val_loss: 0.0854 - val_accuracy: 0.9774 - lr: 0.0010\n",
      "Epoch 35/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1504 - accuracy: 0.9486\n",
      "Epoch 00035: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1504 - accuracy: 0.9486 - val_loss: 0.1009 - val_accuracy: 0.9661 - lr: 0.0010\n",
      "Epoch 36/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1555 - accuracy: 0.9441\n",
      "Epoch 00036: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1555 - accuracy: 0.9441 - val_loss: 0.0934 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 37/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1521 - accuracy: 0.9446\n",
      "Epoch 00037: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1521 - accuracy: 0.9446 - val_loss: 0.1127 - val_accuracy: 0.9616 - lr: 0.0010\n",
      "Epoch 38/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1724 - accuracy: 0.9339\n",
      "Epoch 00038: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1724 - accuracy: 0.9339 - val_loss: 0.0931 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 39/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1452 - accuracy: 0.9469\n",
      "Epoch 00039: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1452 - accuracy: 0.9469 - val_loss: 0.1064 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 40/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1478 - accuracy: 0.9418\n",
      "Epoch 00040: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1478 - accuracy: 0.9418 - val_loss: 0.1270 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 41/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1442 - accuracy: 0.9429\n",
      "Epoch 00041: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1442 - accuracy: 0.9429 - val_loss: 0.0860 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 42/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1508 - accuracy: 0.9384\n",
      "Epoch 00042: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1508 - accuracy: 0.9384 - val_loss: 0.0882 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 43/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "56/55 [==============================] - ETA: 0s - loss: 0.1751 - accuracy: 0.9294\n",
      "Epoch 00043: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1751 - accuracy: 0.9294 - val_loss: 0.0837 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 44/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1617 - accuracy: 0.9356\n",
      "Epoch 00044: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 76s 1s/step - loss: 0.1617 - accuracy: 0.9356 - val_loss: 0.0770 - val_accuracy: 0.9774 - lr: 0.0010\n",
      "Epoch 45/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1329 - accuracy: 0.9475\n",
      "Epoch 00045: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1329 - accuracy: 0.9475 - val_loss: 0.1037 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 46/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1347 - accuracy: 0.9486\n",
      "Epoch 00046: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1347 - accuracy: 0.9486 - val_loss: 0.1285 - val_accuracy: 0.9594 - lr: 0.0010\n",
      "Epoch 47/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1408 - accuracy: 0.9497\n",
      "Epoch 00047: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1408 - accuracy: 0.9497 - val_loss: 0.1067 - val_accuracy: 0.9684 - lr: 0.0010\n",
      "Epoch 48/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1558 - accuracy: 0.9418\n",
      "Epoch 00048: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1558 - accuracy: 0.9418 - val_loss: 0.0796 - val_accuracy: 0.9774 - lr: 0.0010\n",
      "Epoch 49/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1422 - accuracy: 0.9503\n",
      "Epoch 00049: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 69s 1s/step - loss: 0.1422 - accuracy: 0.9503 - val_loss: 0.0944 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 50/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1391 - accuracy: 0.9412\n",
      "Epoch 00050: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1391 - accuracy: 0.9412 - val_loss: 0.1042 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 51/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1428 - accuracy: 0.9480\n",
      "Epoch 00051: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1428 - accuracy: 0.9480 - val_loss: 0.0798 - val_accuracy: 0.9774 - lr: 0.0010\n",
      "Epoch 52/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1445 - accuracy: 0.9435\n",
      "Epoch 00052: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 68s 1s/step - loss: 0.1445 - accuracy: 0.9435 - val_loss: 0.0758 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 53/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1479 - accuracy: 0.9390\n",
      "Epoch 00053: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 93s 2s/step - loss: 0.1479 - accuracy: 0.9390 - val_loss: 0.0779 - val_accuracy: 0.9707 - lr: 0.0010\n",
      "Epoch 54/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1475 - accuracy: 0.9446\n",
      "Epoch 00054: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 77s 1s/step - loss: 0.1475 - accuracy: 0.9446 - val_loss: 0.0851 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 55/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1532 - accuracy: 0.9435\n",
      "Epoch 00055: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 72s 1s/step - loss: 0.1532 - accuracy: 0.9435 - val_loss: 0.1049 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 56/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1289 - accuracy: 0.9525\n",
      "Epoch 00056: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 71s 1s/step - loss: 0.1289 - accuracy: 0.9525 - val_loss: 0.0784 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 57/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1716 - accuracy: 0.9356\n",
      "Epoch 00057: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 72s 1s/step - loss: 0.1716 - accuracy: 0.9356 - val_loss: 0.0793 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 58/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1512 - accuracy: 0.9429\n",
      "Epoch 00058: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 72s 1s/step - loss: 0.1512 - accuracy: 0.9429 - val_loss: 0.1084 - val_accuracy: 0.9639 - lr: 0.0010\n",
      "Epoch 59/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1391 - accuracy: 0.9492\n",
      "Epoch 00059: val_loss did not improve from 0.07471\n",
      "56/55 [==============================] - 72s 1s/step - loss: 0.1391 - accuracy: 0.9492 - val_loss: 0.0848 - val_accuracy: 0.9752 - lr: 0.0010\n",
      "Epoch 60/100\n",
      "56/55 [==============================] - ETA: 0s - loss: 0.1377 - accuracy: 0.9475\n",
      "Epoch 00060: val_loss did not improve from 0.07471\n",
      "Restoring model weights from the end of the best epoch.\n",
      "\n",
      "Epoch 00060: ReduceLROnPlateau reducing learning rate to 0.000800000037997961.\n",
      "56/55 [==============================] - 72s 1s/step - loss: 0.1377 - accuracy: 0.9475 - val_loss: 0.0821 - val_accuracy: 0.9729 - lr: 0.0010\n",
      "Epoch 00060: early stopping\n"
     ]
    }
   ],
   "source": [
    "# assiging the VGG16 model\n",
    "m = VGG19(weights=\"imagenet\", include_top=False, \n",
    "          input_tensor=Input(shape=(64, 64, 3)))\n",
    "\n",
    "\n",
    "model = addTopModel(m, num_classes) \n",
    "\n",
    "for layer in m.layers:\n",
    "    layer.trainable = False\n",
    " \n",
    "\n",
    "    #data augmentation\n",
    "train_generator, validation_generator = process_data(x_train,y_train,\n",
    "                                                 x_test,y_test, batch_size)\n",
    "\n",
    "\n",
    "model.summary()\n",
    "\n",
    "checkpoint = ModelCheckpoint(data_dir+'modelcnp' + '.h5',\n",
    "                             monitor='val_loss',\n",
    "                             mode=\"min\",\n",
    "                             save_best_only=True,\n",
    "                             verbose=1)\n",
    "\n",
    "earlystop = EarlyStopping(monitor='val_loss',\n",
    "                          min_delta=0,\n",
    "                          patience=30,\n",
    "                          verbose=1,\n",
    "                          restore_best_weights=True)\n",
    "\n",
    "learning_rate_reduction = ReduceLROnPlateau(monitor='val_loss',\n",
    "                                            patience=30,\n",
    "                                            verbose=1,\n",
    "                                            factor=0.8,\n",
    "                                            min_lr=0.0001,\n",
    "                                            mode=\"auto\",\n",
    "                                            min_delta=0.0001,\n",
    "                                            cooldown=5)\n",
    "\n",
    "callbacks = [checkpoint, earlystop, learning_rate_reduction]\n",
    "\n",
    "model.compile(loss=models_loss,\n",
    "              optimizer=models_opt,\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "\n",
    "history = model.fit(train_generator,\n",
    "                                epochs=epochs,\n",
    "                              callbacks=callbacks,\n",
    "                              validation_data=validation_generator,\n",
    "                              validation_steps=len(x_test) / batch_size,\n",
    "                              steps_per_epoch=len(x_train) / batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: [0.98245614 0.97426471]\n",
      "Recall Score: [0.96       0.98880597]\n",
      "F1 Score: [0.97109827 0.98148148]\n",
      "Accuracy Score: 0.9774266365688488\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC: 0.9744029850746269\n",
      "Log Loss: 0.07482023782702454\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Confusion matrix\n",
    "plot_confusion_matrix(model, x_test, y_test)\n",
    "## Learning curve\n",
    "plot_learning_curves(history)\n",
    "## ROC AUC\n",
    "multiclass_roc_auc_score(x_test, y_test, model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# saving model\n",
    "model.save(\"modelcnp.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Conclusion\n",
    "The accuracy is 98 % and this is the amount of time the predicted result is actually correct.\n",
    "\n",
    "The recall percentage is 97% and this is the probability of the model diagnosing a correct positive diagnosis out of all the times it diagnosed positive. This would be the best metric in this case as we would rather give a wrong positive diagnosis than give a wrong negative diagnosis.\n",
    "\n",
    "### Recommendation\n",
    "* Reshape the x-ray sizes to 64 x 64 to reduce the amount of computational time and power consumed when using the VGG19 model.\n",
    "\n",
    "\n",
    "* Use a maxpooling layer before flatten the model and using a dense layer of 512 nodes and also a learning rate of 0.000800000037997961 in order to get a better performing model when using the VGG19 model.\n",
    "\n",
    "\n",
    "* Add a dropout layer before the final dense layer to dropout half of the output from the prior dense layer using 512 nodes in order to reduce overfitting when using the VGG19 model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "authorship_tag": "ABX9TyOBK/qqRk5TcvVbe4on/F9j",
   "collapsed_sections": [
    "CIQEU2HjTPzF",
    "4Zlv5kGh3LvC",
    "71bunMrmSEGF",
    "7TopSN6fOcmf",
    "ZyYB-p4qOTNx",
    "_wBgbcDEOqJd",
    "wYCX523UOW5K",
    "TalwgYalOdPd",
    "Cf2dxyj6O3yt",
    "BXzugA8pOdlv"
   ],
   "include_colab_link": true,
   "mount_file_id": "1eAzBe9FO37HcJfumZGrqDBb7IyNw7AuW",
   "name": "v7_3 classes_bin_covid19_multiple_models.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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}