CODE_model2_train.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "IoGS1RK1v_jI"
   },
   "outputs": [],
   "source": [
    "# 1st Step – Import the libraries that will be used\n",
    "\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.applications import VGG16\n",
    "from tensorflow.keras.applications import Xception\n",
    "from tensorflow.keras.applications import ResNet50V2\n",
    "from tensorflow.keras.layers import AveragePooling2D\n",
    "from tensorflow.keras.layers import MaxPooling2D\n",
    "from tensorflow.keras.layers import Dropout\n",
    "from tensorflow.keras.layers import Flatten\n",
    "from tensorflow.keras.layers import Conv2D\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.callbacks import ReduceLROnPlateau\n",
    "from tensorflow.keras.layers import Input\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from imutils import paths\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import argparse\n",
    "import cv2\n",
    "import os\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import pandas as pd\n",
    "import seaborn as sn\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "CXwTGiPpwg-s"
   },
   "outputs": [],
   "source": [
    "# 2nd Step – Load the arrays built in the step referring to data pre-processing and normalize the input data\n",
    "\n",
    "X_train = np.load(\"/content/drive/My Drive/Python/COVID/Arrays/Modelo2/X_Train.npy\")\n",
    "X_train = X_train/255\n",
    "\n",
    "Y_train = np.load(\"/content/drive/My Drive/Python/COVID/Arrays/Modelo2/Y_Train.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 726,
     "status": "ok",
     "timestamp": 1587496628086,
     "user": {
      "displayName": "César Soares",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GjnCuTGeBsi0P9nLHQXzmAvmSU4ekinz51pT1mltQ=s64",
      "userId": "03845423099530420401"
     },
     "user_tz": 180
    },
    "id": "1RKojXsDxAkB",
    "outputId": "adb8a407-c862-4250-a7d1-5171add8ceaf"
   },
   "outputs": [],
   "source": [
    "# 3rd Step – Split the data into training data and test data\n",
    "\n",
    "X_train,X_test,Y_train,Y_test = train_test_split(X_train,Y_train, test_size = 0.2, random_state = 40)\n",
    "\n",
    "print(f\"X_train shape: {X_train.shape} Y_train shape {Y_train.shape}\")\n",
    "print(f\"X_test shape: {X_test.shape} Y_test shape {Y_test.shape}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "T9b0A-R_yPvu"
   },
   "outputs": [],
   "source": [
    "# 4th Step – Determining the architecture of the model (Xception) that will be trained\n",
    "\n",
    "bModel = Xception(weights=\"imagenet\", include_top=False,\n",
    "        input_tensor=Input(shape=(237, 237, 3)))\n",
    "tModel = bModel.output\n",
    "tModel = AveragePooling2D(pool_size=(2, 2))(tModel)\n",
    "tModel = Flatten(name=\"flatten\")(tModel)\n",
    "tModel = Dense(20, activation=\"relu\")(tModel)\n",
    "tModel = Dropout(0.2)(tModel)\n",
    "tModel = Dense(3, activation=\"softmax\")(tModel)\n",
    "\n",
    "model = Model(inputs=bModel.input, outputs=tModel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 5th Step – Determine the hyperparameters and compile the model (Xception)\n",
    "\n",
    "INIT_LR = 1e-3\n",
    "EPOCHS = 80\n",
    "BS = 15\n",
    "\n",
    "for layer in bModel.layers:\n",
    "    layer.trainable = False\n",
    "\n",
    "opt = Adam(lr=INIT_LR)\n",
    "model.compile(loss=\"categorical_crossentropy\", optimizer=opt,\n",
    "    metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 6th Step – Train the model (Xception)\n",
    "\n",
    "from tensorflow.keras.callbacks import ReduceLROnPlateau\n",
    "reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2,\n",
    "                              patience=5, min_lr=0.001, cooldown=5)\n",
    "\n",
    "x = model.fit(X_train, Y_train, batch_size=BS,validation_split=0.1, epochs=EPOCHS,callbacks=[reduce_lr])\n",
    "\n",
    "model.save(\"/content/drive/My Drive/Python/COVID/model/modeloc_2.hdf5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 7th Step - Observe the accuracy of the model (Xception) and the loss function\n",
    "\n",
    "plt.plot(x.history['accuracy'])\n",
    "plt.plot(x.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(x.history['loss'])\n",
    "plt.plot(x.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "model.evaluate(X_test,Y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 8th Step – Determining the architecture of the model (ResNet50V2) that will be trained\n",
    "\n",
    "bModel = ResNet50V2(weights=\"imagenet\", include_top=False,\n",
    "        input_tensor=Input(shape=(237, 237, 3)))\n",
    "tModel = bModel.output\n",
    "tModel = AveragePooling2D(pool_size=(2, 2))(tModel)\n",
    "tModel = Flatten(name=\"flatten\")(tModel)\n",
    "tModel = Dense(20, activation=\"relu\")(tModel)\n",
    "tModel = Dropout(0.2)(tModel)\n",
    "tModel = Dense(3, activation=\"softmax\")(tModel)\n",
    "\n",
    "model = Model(inputs=bModel.input, outputs=tModel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 9th Step – Determine the hyperparameters and compile the model (ResNet50V2)\n",
    "\n",
    "INIT_LR = 1e-3\n",
    "EPOCHS = 80\n",
    "BS = 15\n",
    "\n",
    "for layer in bModel.layers:\n",
    "\tlayer.trainable = False\n",
    "\n",
    "opt = Adam(lr=INIT_LR)\n",
    "model.compile(loss=\"categorical_crossentropy\", optimizer=opt,\n",
    "\tmetrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 10th Step – Train the model (ResNet50V2)\n",
    "\n",
    "from tensorflow.keras.callbacks import ReduceLROnPlateau\n",
    "reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2,\n",
    "                              patience=5, min_lr=0.001, cooldown=5)\n",
    "\n",
    "x = model.fit(X_train, Y_train, batch_size=BS,validation_split=0.1, epochs=EPOCHS,callbacks=[reduce_lr])\n",
    "\n",
    "model.save(\"/content/drive/My Drive/Python/COVID/model/modeloc_2.hdf5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 11th Step - Observe the accuracy of the model (ResNet50V2) and the loss function\n",
    "\n",
    "plt.plot(x.history['accuracy'])\n",
    "plt.plot(x.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(x.history['loss'])\n",
    "plt.plot(x.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "model.evaluate(X_test,Y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 12º Step – Determining the architecture (VGG16) of the model that will be trained\n",
    "\n",
    "bModel = VGG16(weights=\"imagenet\", include_top=False,classes=3,\n",
    "        input_tensor=Input(shape=(237, 237, 3)))\n",
    "tModel = bModel.output\n",
    "tModel = AveragePooling2D(pool_size=(2, 2))(tModel)\n",
    "tModel = Flatten(name=\"flatten\")(tModel)\n",
    "tModel = Dense(20, activation=\"relu\")(tModel)\n",
    "tModel = Dropout(0.2)(tModel)\n",
    "tModel = Dense(3, activation=\"softmax\")(tModel)\n",
    "\n",
    "model = Model(inputs=bModel.input, outputs=tModel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "bMDJX0paylb4"
   },
   "outputs": [],
   "source": [
    "# 13º Step – Determine the hyperparameters and compile the model (VGG16)\n",
    "\n",
    "INIT_LR = 1e-3\n",
    "EPOCHS = 80\n",
    "BS = 15\n",
    "\n",
    "for layer in bModel.layers:\n",
    "\tlayer.trainable = False\n",
    "\n",
    "opt = Adam(lr=INIT_LR)\n",
    "model.compile(loss=\"categorical_crossentropy\", optimizer=opt,\n",
    "\tmetrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 96997,
     "status": "ok",
     "timestamp": 1587496742305,
     "user": {
      "displayName": "César Soares",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GjnCuTGeBsi0P9nLHQXzmAvmSU4ekinz51pT1mltQ=s64",
      "userId": "03845423099530420401"
     },
     "user_tz": 180
    },
    "id": "fXEHkwQ4y-bw",
    "outputId": "c749a8d2-9930-44ba-d265-cba28f25d372"
   },
   "outputs": [],
   "source": [
    "# 14th Step – Train the model (VGG16)\n",
    "\n",
    "reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2,\n",
    "                              patience=5, min_lr=0.001, cooldown=5)\n",
    "\n",
    "x = model.fit(X_train, Y_train, batch_size=BS,validation_split=0.1, epochs=EPOCHS,callbacks=[reduce_lr])\n",
    "\n",
    "model.save(\"/content/drive/My Drive/Python/COVID/model/modeloc_2.hdf5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 607
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 2344,
     "status": "ok",
     "timestamp": 1587496758950,
     "user": {
      "displayName": "César Soares",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GjnCuTGeBsi0P9nLHQXzmAvmSU4ekinz51pT1mltQ=s64",
      "userId": "03845423099530420401"
     },
     "user_tz": 180
    },
    "id": "s996fVhN0Bba",
    "outputId": "c89b6c22-1641-4b2c-b111-99823f4cefb2"
   },
   "outputs": [],
   "source": [
    "# 15th Step – Observe the accuracy of the model (VGG16) and the loss function\n",
    "\n",
    "plt.plot(x.history['accuracy'])\n",
    "plt.plot(x.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(x.history['loss'])\n",
    "plt.plot(x.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "model.evaluate(X_test,Y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 8527,
     "status": "ok",
     "timestamp": 1587398056105,
     "user": {
      "displayName": "César Soares",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GjnCuTGeBsi0P9nLHQXzmAvmSU4ekinz51pT1mltQ=s64",
      "userId": "03845423099530420401"
     },
     "user_tz": 180
    },
    "id": "ONo53UaLXq2U",
    "outputId": "4369dd70-e75c-419a-900a-88a526cfb7ac"
   },
   "outputs": [],
   "source": [
    "# 16º Step – Observe which images the model (VGG16) got right\n",
    "\n",
    "plt.figure(figsize=(20,20))\n",
    "plt.subplots_adjust(left=None, bottom=None, right=None, top=None, wspace=2.0, hspace=2.0)\n",
    "i = 0\n",
    "for i,image in enumerate(X_test):\n",
    "    plt.subplot(9,9,i+1)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.grid(False)\n",
    "    plt.imshow(image, cmap=plt.cm.binary)\n",
    "    img = np.expand_dims(X_test[i],axis = 0)\n",
    "    x_pred = model.predict(img)[0]\n",
    "    pred_covid = x_pred[0]\n",
    "   \n",
    "    pred_normal = x_pred[1]\n",
    "\n",
    "    pred_infeccoes = x_pred[2]\n",
    "    \n",
    "    \n",
    "    if pred_covid > pred_normal and pred_covid > pred_infeccoes:\n",
    "      label = \"COVID\"\n",
    "    elif pred_normal > pred_covid and pred_normal > pred_infeccoes:\n",
    "      label = \"NORMAL\"\n",
    "    elif pred_infeccoes > pred_covid and pred_infeccoes > pred_normal:\n",
    "      label = \"INFECÇÕES\"\n",
    "     \n",
    "    \n",
    "    if Y_test[i][0] == 1:\n",
    "      label_test = \"COVID\"\n",
    "    elif Y_test[i][1] == 1:\n",
    "      label_test = \"NORMAL\"\n",
    "    elif Y_test[i][2] == 1:\n",
    "      label_test = \"INFECÇÕES\"\n",
    "    plt.xlabel(f\"Label Predict = {label} \\n Label Correct = {label_test}\")\n",
    "    i += 1\n",
    "plt.savefig('/content/drive/My Drive/Python/COVID/model/modelo_2.pdf')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "q-6ORBcKgQEd"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "treinamento_resultados_COVID_modelo2.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}