Imagenet_CNN_v3.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Imagenet CNN v2.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "lGg6HhWrqVgb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install -U -q kaggle\n",
        "!mkdir -p ~/.kaggle\n",
        "from google.colab import files\n",
        "import os\n",
        "%tensorflow_version 1.x\n",
        "from tensorflow.keras.applications.inception_v3 import InceptionV3\n",
        "from tensorflow.keras.preprocessing import image\n",
        "from tensorflow.keras.models import Model, load_model\n",
        "from tensorflow.keras.callbacks import ModelCheckpoint, ReduceLROnPlateau\n",
        "from tensorflow.keras.layers import Dense, GlobalAveragePooling2D, Dropout\n",
        "from tensorflow.keras.layers import BatchNormalization, Activation\n",
        "from tensorflow.python.keras.optimizers import Adam, RMSprop\n",
        "from tensorflow.python.keras.preprocessing.image import ImageDataGenerator\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EQcKSnGOrfjc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "files.upload()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lQTzZP6d3lzL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "image_size = 299\n",
        "batch_size = 32"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u786b-yjriOc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!cp kaggle.json ~/.kaggle/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "371vEuiU0Gg5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!kaggle datasets list"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LB6CJTT90W7U",
        "colab_type": "text"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SOskQsVm0Y4q",
        "colab_type": "text"
      },
      "source": [
        "kaggle datasets download -d paultimothymooney/chest-xray-pneumonia"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DVIo4SmV0bXy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!kaggle datasets download -d paultimothymooney/chest-xray-pneumonia"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uRnTZLUq1LWC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!chmod 600 /root/.kaggle/kaggle.json"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jyO8w-CYroOF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!ls"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b3ZeGRsu0V1z",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!unzip chest-xray-pneumonia.zip"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2k6bOApW4Q9y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!ls chest_xray/val/PNEUMONIA -1 | wc -l"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jVxVYnqJT5oy",
        "colab_type": "text"
      },
      "source": [
        "**TEST:**: NORMAL — 234, PNEUMONIA — 390\n",
        "\n",
        "**TRAIN:** NORMAL — 1341, PNEUMONIA — 3875\n",
        "\n",
        "**VALIDATION:** NORMAL — 8, PNEUMONIA — 8\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Gzg3gheEYUn8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_dir = 'chest_xray/train'\n",
        "test_dir = 'chest_xray/test'\n",
        "input_shape = (image_size,image_size,3)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L4KNqrdmY2BL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_datagen = ImageDataGenerator(rescale=1. / 255,\n",
        "    rotation_range=10,\n",
        "    width_shift_range=0.1,\n",
        "    height_shift_range=0.1,\n",
        "    zoom_range=0.1,\n",
        "    fill_mode='nearest')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CSE2ac5BeMt1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "img_path = 'chest_xray/test/NORMAL/NORMAL2-IM-0380-0001.jpeg'\n",
        "img = image.load_img(img_path, target_size=(image_size, image_size))\n",
        "x = image.img_to_array(img)\n",
        "x = x.reshape((1,) + x.shape)\n",
        "i = 0\n",
        "for batch in train_datagen.flow(x, batch_size=1):\n",
        "    plt.figure(i)\n",
        "    imgplot = plt.imshow(image.array_to_img(batch[0]))\n",
        "    i += 1\n",
        "    if i==8:\n",
        "      break;\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j5dmxIFMiBqO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_generator = train_datagen.flow_from_directory(\n",
        "    train_dir,\n",
        "    target_size=(image_size, image_size),\n",
        "    batch_size=batch_size,\n",
        "    class_mode='binary')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H4dgTl05jYNY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_datagen = ImageDataGenerator(rescale=1. / 255)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P3tfPsXjjYYS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_generator = test_datagen.flow_from_directory(\n",
        "    test_dir,\n",
        "    target_size=(image_size, image_size),\n",
        "    batch_size=batch_size,\n",
        "    class_mode='binary')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VnYlIcJJiedc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pretrained_model = InceptionV3(weights='imagenet', include_top=False)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MmslochLirKf",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "x = pretrained_model.output\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "x = Dense(512)(x)\n",
        "x = BatchNormalization()(x)\n",
        "x = Activation('relu')(x)\n",
        "x = Dropout(0.5)(x)\n",
        "predictions = Dense(1, activation='sigmoid')(x)\n",
        "model = Model(inputs=pretrained_model.input, outputs=predictions)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ojKQVofri-kf",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer = RMSprop(lr = 1e-4), \n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "05pYapfbjAbm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "filepath=\"inceptionv3_best.h5\"\n",
        "checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, \n",
        "                             save_best_only=True, mode='max')\n",
        "learning_rate_reduction = ReduceLROnPlateau(monitor='val_acc', \n",
        "                                            patience=3, \n",
        "                                            verbose=1, \n",
        "                                            factor=0.5, \n",
        "                                            min_lr=0.00001)\n",
        "callbacks_list = [checkpoint, learning_rate_reduction]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ExLM5Ws_jF8y",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "steps_per_epoch = 40\n",
        "history = model.fit_generator(\n",
        "    train_generator,\n",
        "    steps_per_epoch=steps_per_epoch,\n",
        "    validation_data=test_generator,\n",
        "    validation_steps=5,\n",
        "    callbacks=callbacks_list,\n",
        "    epochs=20,\n",
        "    verbose=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i3zIl9aAqzBG",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.load_weights(\"inceptionv3_best.h5\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ggUjUPRsm7ir",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pretrained_model.trainable = False\n",
        "for layer in model.layers[:249]:\n",
        "   layer.trainable = False\n",
        "for layer in model.layers[249:]:\n",
        "   layer.trainable = True"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8O3vCFBQnPWu",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(loss='binary_crossentropy',\n",
        "              optimizer=RMSprop(lr=1e-4), \n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ox4jeeWEng0M",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "filepath=\"inceptionv3_fine_tuned.h5\"\n",
        "checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
        "callbacks_list = [checkpoint, learning_rate_reduction]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MKZd4SXlniiP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "steps_per_epoch = 40\n",
        "history = model.fit_generator(\n",
        "    train_generator,\n",
        "    steps_per_epoch=steps_per_epoch,\n",
        "    validation_data=test_generator,\n",
        "    validation_steps=5,\n",
        "    callbacks=callbacks_list,\n",
        "    epochs=5,\n",
        "    verbose=2)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WzA5VqnrnlYn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = load_model(\"inceptionv3_fine_tuned.h5\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DLKkY054oXQA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "plt.plot(history.history['acc'])\n",
        "plt.plot(history.history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0VSwzc_Gorxz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.metrics import classification_report, confusion_matrix\n",
        "import itertools"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uRZ2luixpOB-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "val_dir = 'chest_xray/val'\n",
        "val_datagen = image.ImageDataGenerator(rescale=1. / 255)\n",
        "val_generator = val_datagen.flow_from_directory(\n",
        "    val_dir,\n",
        "    target_size=(image_size, image_size),\n",
        "    batch_size=batch_size,\n",
        "    class_mode='binary')\n",
        "true_res = list()\n",
        "predicted_res = list()\n",
        "\n",
        "for i in val_generator.filenames:\n",
        "    path = i.split(\"/\")\n",
        "    if (path[0] == \"NORMAL\"):\n",
        "      true_res.append(0)\n",
        "    else:\n",
        "      true_res.append(1)\n",
        "    path_im = \"chest_xray/val/\" + i\n",
        "    img = image.load_img(path_im, target_size=(image_size, image_size))\n",
        "    plt.imshow(img)\n",
        "    plt.title(path[0])\n",
        "    plt.show()  \n",
        "    x = image.img_to_array(img)\n",
        "    x /=255\n",
        "    x = np.expand_dims(x, axis=0)\n",
        "    prediction = model.predict(x)\n",
        "    print(prediction)\n",
        "    if (prediction <=0.5):\n",
        "      predicted_res.append(0)\n",
        "    else:\n",
        "      predicted_res.append(1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Csn6Nmb9pyh5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def plot_confusion_matrix(cm, classes,\n",
        "                          normalize=False,\n",
        "                          title='Confusion matrix',\n",
        "                          cmap=plt.cm.Blues):\n",
        "    \"\"\"\n",
        "    This function prints and plots the confusion matrix.\n",
        "    Normalization can be applied by setting `normalize=True`.\n",
        "    \"\"\"\n",
        "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
        "    plt.title(title)\n",
        "    plt.colorbar()\n",
        "    tick_marks = np.arange(len(classes))\n",
        "    plt.xticks(tick_marks, classes, rotation=45)\n",
        "    plt.yticks(tick_marks, classes)\n",
        "\n",
        "    if normalize:\n",
        "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
        "        print(\"Normalized confusion matrix\")\n",
        "    else:\n",
        "        print('Confusion matrix, without normalization')\n",
        "\n",
        "    print(cm)\n",
        "\n",
        "    thresh = cm.max() / 2.\n",
        "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
        "        plt.text(j, i, cm[i, j],\n",
        "                 horizontalalignment=\"center\",\n",
        "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
        "\n",
        "    plt.tight_layout()\n",
        "    plt.ylabel('True label')\n",
        "    plt.xlabel('Predicted label')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lI3z-IAQpTBb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "font = {'size' : 15}\n",
        "\n",
        "plt.rc('font', **font)\n",
        "\n",
        "cnf_matrix = confusion_matrix(true_res,predicted_res)\n",
        "plt.figure(figsize=(10, 8))\n",
        "plot_confusion_matrix(cnf_matrix, classes=[\"NORMAL\", \"PNEUMONIA\"],\n",
        "                      title='Confusion matrix')\n",
        "plt.savefig(\"conf_matrix.png\")\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QLICKeO9pxz0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jHjaiKPGpdcy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}