best_cnn_segmentation_connected_components_384x384_2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "igEd7obtcxR2"
   },
   "source": [
    "# Approach\n",
    "\n",
    "* Firstly a convolutional neural network is used to segment the image, using the bounding boxes directly as a mask. \n",
    "* Secondly connected components is used to separate multiple areas of predicted pneumonia.\n",
    "* Finally a bounding box is simply drawn around every connected component.\n",
    "\n",
    "# Network\n",
    "\n",
    "* The network consists of a number of residual blocks with convolutions and downsampling blocks with max pooling.\n",
    "* At the end of the network a single upsampling layer converts the output to the same shape as the input.\n",
    "\n",
    "As the input to the network is 256 by 256 (instead of the original 1024 by 1024) and the network downsamples a number of times without any meaningful upsampling (the final upsampling is just to match in 256 by 256 mask) the final prediction is very crude. If the network downsamples 4 times the final bounding boxes can only change with at least 16 pixels.\n",
    "\n",
    "**Edit by EAS** - Change input image size to 320x320."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2018-09-08\n"
     ]
    }
   ],
   "source": [
    "import datetime\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vlU7J4Uvc9jY"
   },
   "outputs": [],
   "source": [
    "# install dependencies not included by Colab\n",
    "# use pip3 to ensure compatibility w/ Google Deep Learning Images \n",
    "!pip install -q pydicom \n",
    "!pip install -q imgaug \n",
    "!pip install -q kaggle\n",
    "!pip install -q tqdm \n",
    "!pip install -q PyDrive"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LHncOor-cxSS"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\eric\\anaconda2\\envs\\hse_dl\\lib\\site-packages\\h5py\\__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import csv\n",
    "import random\n",
    "import pydicom\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from skimage import measure\n",
    "from skimage.transform import resize\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "xu0hwg-MdA1l"
   },
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'google.colab'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-2-da40f1c46ddf>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[1;32mfrom\u001b[0m \u001b[0mgoogle\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolab\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mauth\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0mgoogleapiclient\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mhttp\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mMediaFileUpload\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0mgoogleapiclient\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdiscovery\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mbuild\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mauth\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mauthenticate_user\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mModuleNotFoundError\u001b[0m: No module named 'google.colab'"
     ]
    }
   ],
   "source": [
    "from google.colab import auth\n",
    "from googleapiclient.http import MediaFileUpload\n",
    "from googleapiclient.discovery import build\n",
    "\n",
    "auth.authenticate_user()\n",
    "\n",
    "drive_service = build('drive', 'v3')\n",
    "\n",
    "def save_file_to_drive(name, path):\n",
    "    file_metadata = {\n",
    "      'name': name,\n",
    "      'mimeType': 'application/octet-stream'\n",
    "     }\n",
    "\n",
    "    media = MediaFileUpload(path, \n",
    "                    mimetype='application/octet-stream',\n",
    "                    resumable=True)\n",
    "\n",
    "    created = drive_service.files().create(body=file_metadata,\n",
    "                                   media_body=media,\n",
    "                                   fields='id').execute()\n",
    "\n",
    "    print('File ID: {}'.format(created.get('id')))\n",
    "\n",
    "    return created"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hE3jDmgudDDP"
   },
   "outputs": [],
   "source": [
    "# enter your Kaggle credentionals here\n",
    "os.environ['KAGGLE_USERNAME']=\"skooch\"\n",
    "os.environ['KAGGLE_KEY']=\"42f8a02ee92cc773d1dbe66565673ad3\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 352
    },
    "colab_type": "code",
    "id": "jowYWnxzdDtf",
    "outputId": "8a7a5976-8615-4a08-cca5-f25a64fd2163"
   },
   "outputs": [],
   "source": [
    "# If you are unable to download the competition dataset, check to see if you have \n",
    "# accepted the user agreement on the competition website. \n",
    "if not os.path.exists(\"./stage_1_detailed_class_info.csv.zip\"):\n",
    "    !kaggle competitions download -c rsna-pneumonia-detection-challenge"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 36
    },
    "colab_type": "code",
    "id": "HttYLaczdKlJ",
    "outputId": "bd3fb4d1-e954-4c29-b69f-a305579b7ad2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unzipping...\n"
     ]
    }
   ],
   "source": [
    "# unzipping takes a few minutes\n",
    "if not os.path.exists(\"./stage_1_train_labels.csv\"):\n",
    "    print(\"Unzipping...\")\n",
    "    !unzip -q -o stage_1_test_images.zip -d stage_1_test_images\n",
    "    !unzip -q -o stage_1_train_images.zip -d stage_1_train_images\n",
    "    !unzip -q -o stage_1_train_labels.csv.zip"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "sSWY-U2TcxSn"
   },
   "source": [
    "# Load pneumonia locations\n",
    "\n",
    "Table contains [filename : pneumonia location] pairs per row. \n",
    "* If a filename contains multiple pneumonia, the table contains multiple rows with the same filename but different pneumonia locations. \n",
    "* If a filename contains no pneumonia it contains a single row with an empty pneumonia location.\n",
    "\n",
    "The code below loads the table and transforms it into a dictionary. \n",
    "* The dictionary uses the filename as key and a list of pneumonia locations in that filename as value. \n",
    "* If a filename is not present in the dictionary it means that it contains no pneumonia."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 36
    },
    "colab_type": "code",
    "id": "m_0ZGQyLdQS8",
    "outputId": "3e6787f3-971e-4fe4-e12d-c5073d51175f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train dir: ./stage_1_train_images\n"
     ]
    }
   ],
   "source": [
    "ROOT_DIR = \"./\"\n",
    "\n",
    "train_dicom_dir = os.path.join(ROOT_DIR, 'stage_1_train_images')\n",
    "test_dicom_dir = os.path.join(ROOT_DIR, 'stage_1_test_images')\n",
    "print(\"Train dir:\", train_dicom_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hopsRmICcxSs"
   },
   "outputs": [],
   "source": [
    "# empty dictionary\n",
    "pneumonia_locations = {}\n",
    "# load table\n",
    "with open(os.path.join('./stage_1_train_labels.csv'), mode='r') as infile:\n",
    "    # open reader\n",
    "    reader = csv.reader(infile)\n",
    "    # skip header\n",
    "    next(reader, None)\n",
    "    # loop through rows\n",
    "    for rows in reader:\n",
    "        # retrieve information\n",
    "        filename = rows[0]\n",
    "        location = rows[1:5]\n",
    "        pneumonia = rows[5]\n",
    "        # if row contains pneumonia add label to dictionary\n",
    "        # which contains a list of pneumonia locations per filename\n",
    "        if pneumonia == '1':\n",
    "            # convert string to float to int\n",
    "            location = [int(float(i)) for i in location]\n",
    "            # save pneumonia location in dictionary\n",
    "            if filename in pneumonia_locations:\n",
    "                pneumonia_locations[filename].append(location)\n",
    "            else:\n",
    "                pneumonia_locations[filename] = [location]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "UoB6GBhIcxS6"
   },
   "source": [
    "# Load filenames"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "colab_type": "code",
    "id": "b_iB-IftcxTE",
    "outputId": "25259202-5cfe-4b1b-ef72-dea97128ce53"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n train samples 477\n",
      "n valid samples 53\n"
     ]
    }
   ],
   "source": [
    "# load and shuffle filenames\n",
    "folder = './stage_1_train_images'\n",
    "filenames = os.listdir(folder)\n",
    "random.shuffle(filenames)\n",
    "# split into train and validation filenames\n",
    "n_valid_samples = int(len(filenames) * 0.1)\n",
    "train_filenames = filenames[n_valid_samples:]\n",
    "valid_filenames = filenames[:n_valid_samples]\n",
    "print('n train samples', len(train_filenames))\n",
    "print('n valid samples', len(valid_filenames))\n",
    "n_train_samples = len(filenames) - n_valid_samples"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "lorV7ZmDcxTa"
   },
   "source": [
    " # Data generator\n",
    "\n",
    "The dataset is too large to fit into memory, so we need to create a generator that loads data on the fly.\n",
    "\n",
    "* The generator takes in some filenames, batch_size and other parameters.\n",
    "\n",
    "* The generator outputs a random batch of numpy images and numpy masks.\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "dFbLdFxVehhB"
   },
   "outputs": [],
   "source": [
    "BATCH_SIZE = 12\n",
    "IMAGE_SIZE = 448\n",
    "CHECKPOINT_PATH = \"model6_448.h5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TueY1bVlcxTg"
   },
   "outputs": [],
   "source": [
    "class generator(keras.utils.Sequence):\n",
    "    \n",
    "    def __init__(self, folder, filenames, pneumonia_locations=None, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=True, augment=False, predict=False):\n",
    "        self.folder = folder\n",
    "        self.filenames = filenames\n",
    "        self.pneumonia_locations = pneumonia_locations\n",
    "        self.batch_size = batch_size\n",
    "        self.image_size = image_size\n",
    "        self.shuffle = shuffle\n",
    "        self.augment = augment\n",
    "        self.predict = predict\n",
    "        self.on_epoch_end()\n",
    "        \n",
    "    def __load__(self, filename):\n",
    "        # load dicom file as numpy array\n",
    "        img = pydicom.dcmread(os.path.join(self.folder, filename)).pixel_array\n",
    "        # create empty mask\n",
    "        msk = np.zeros(img.shape)\n",
    "        # get filename without extension\n",
    "        filename = filename.split('.')[0]\n",
    "        # if image contains pneumonia\n",
    "        if filename in pneumonia_locations:\n",
    "            # loop through pneumonia\n",
    "            for location in pneumonia_locations[filename]:\n",
    "                # add 1's at the location of the pneumonia\n",
    "                x, y, w, h = location\n",
    "                msk[y:y+h, x:x+w] = 1\n",
    "        # if augment then horizontal flip half the time\n",
    "        if self.augment and random.random() > 0.5:\n",
    "            img = np.fliplr(img)\n",
    "            msk = np.fliplr(msk)\n",
    "        # resize both image and mask\n",
    "        img = resize(img, (self.image_size, self.image_size), mode='reflect')\n",
    "        msk = resize(msk, (self.image_size, self.image_size), mode='reflect') > 0.5\n",
    "        # add trailing channel dimension\n",
    "        img = np.expand_dims(img, -1)\n",
    "        msk = np.expand_dims(msk, -1)\n",
    "        return img, msk\n",
    "    \n",
    "    def __loadpredict__(self, filename):\n",
    "        # load dicom file as numpy array\n",
    "        img = pydicom.dcmread(os.path.join(self.folder, filename)).pixel_array\n",
    "        # resize image\n",
    "        img = resize(img, (self.image_size, self.image_size), mode='reflect')\n",
    "        # add trailing channel dimension\n",
    "        img = np.expand_dims(img, -1)\n",
    "        return img\n",
    "        \n",
    "    def __getitem__(self, index):\n",
    "        # select batch\n",
    "        filenames = self.filenames[index*self.batch_size:(index+1)*self.batch_size]\n",
    "        # predict mode: return images and filenames\n",
    "        if self.predict:\n",
    "            # load files\n",
    "            imgs = [self.__loadpredict__(filename) for filename in filenames]\n",
    "            # create numpy batch\n",
    "            imgs = np.array(imgs)\n",
    "            return imgs, filenames\n",
    "        # train mode: return images and masks\n",
    "        else:\n",
    "            # load files\n",
    "            items = [self.__load__(filename) for filename in filenames]\n",
    "            # unzip images and masks\n",
    "            imgs, msks = zip(*items)\n",
    "            # create numpy batch\n",
    "            imgs = np.array(imgs)\n",
    "            msks = np.array(msks)\n",
    "            return imgs, msks\n",
    "        \n",
    "    def on_epoch_end(self):\n",
    "        if self.shuffle:\n",
    "            random.shuffle(self.filenames)\n",
    "        \n",
    "    def __len__(self):\n",
    "        if self.predict:\n",
    "            # return everything\n",
    "            return int(np.ceil(len(self.filenames) / self.batch_size))\n",
    "        else:\n",
    "            # return full batches only\n",
    "            return int(len(self.filenames) / self.batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3_GgDL7ncxT3"
   },
   "source": [
    "# Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "bqobZQZQcxUE"
   },
   "outputs": [],
   "source": [
    "def create_downsample(channels, inputs):\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(inputs)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "#     x = keras.layers.Conv2D(channels, 1, padding='same', use_bias=False)(x)\n",
    "#     x = keras.layers.MaxPool2D(2)(x)\n",
    "    x = keras.layers.Conv2D(channels, (3,3), strides=(2,2), padding='same', use_bias=False)(x)\n",
    "    return x\n",
    "\n",
    "def create_resblock(channels, inputs):\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(inputs)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(x)\n",
    "    return keras.layers.add([x, inputs])\n",
    "\n",
    "def create_network(input_size, channels, n_blocks=2, depth=4):\n",
    "    # input\n",
    "    inputs = keras.Input(shape=(input_size, input_size, 1))\n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(inputs)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    x = keras.layers.Conv2D(channels, 3, padding='same', use_bias=False)(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    x = keras.layers.Conv2D(channels, 3, strides=(2,2), padding='same', use_bias=False)(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    x = keras.layers.Conv2D(64, 1, padding='same', use_bias=False)(x)\n",
    "    \n",
    "    # residual blocks\n",
    "    for d in range(depth):\n",
    "        channels = channels * 2\n",
    "        if d > 0:\n",
    "            x = create_downsample(channels, x)\n",
    "            \n",
    "        for b in range(n_blocks):\n",
    "            x = create_resblock(channels, x)\n",
    "        \n",
    "    # output\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    x_2 = keras.layers.Conv2D(512, (3,3), padding='same', dilation_rate=(2,2), activation=None, name=\"dilated_conv_1\")(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x_2)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    x = keras.layers.Conv2D(512, (3,3), padding='same', dilation_rate=(2,2), activation=None, name=\"dilated_conv_2\")(x)\n",
    "    # residual connection\n",
    "    x = keras.layers.add([x, x_2])\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    # reduce number of feature maps\n",
    "    x = keras.layers.Conv2D(384, 1, activation=None)(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    # transpose convolution to upsize\n",
    "    x = keras.layers.Conv2DTranspose(256, (8,8), (4,4), padding=\"same\", activation=None)(x)\n",
    "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
    "    x = keras.layers.LeakyReLU(0)(x)\n",
    "    \n",
    "    x = keras.layers.Conv2D(1, 1, activation='sigmoid')(x)\n",
    "    outputs = keras.layers.UpSampling2D(2**(depth-2))(x)\n",
    "    \n",
    "    model = keras.Model(inputs=inputs, outputs=outputs)\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "AvdSzGI4cxUL"
   },
   "source": [
    "# Train network\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 3666
    },
    "colab_type": "code",
    "id": "DCgIz8CwcxUO",
    "outputId": "babc5e06-7bb4-42f1-a742-e29d57d8b5c3"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_10 (InputLayer)           (None, 448, 448, 1)  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_196 (Conv2D)             (None, 448, 448, 32) 288         input_10[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_193 (BatchN (None, 448, 448, 32) 128         conv2d_196[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_193 (LeakyReLU)     (None, 448, 448, 32) 0           batch_normalization_193[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_197 (Conv2D)             (None, 448, 448, 32) 9216        leaky_re_lu_193[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_194 (BatchN (None, 448, 448, 32) 128         conv2d_197[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_194 (LeakyReLU)     (None, 448, 448, 32) 0           batch_normalization_194[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_198 (Conv2D)             (None, 224, 224, 32) 9216        leaky_re_lu_194[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_195 (BatchN (None, 224, 224, 32) 128         conv2d_198[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_195 (LeakyReLU)     (None, 224, 224, 32) 0           batch_normalization_195[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_199 (Conv2D)             (None, 224, 224, 64) 2048        leaky_re_lu_195[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_196 (BatchN (None, 224, 224, 64) 256         conv2d_199[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_196 (LeakyReLU)     (None, 224, 224, 64) 0           batch_normalization_196[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_200 (Conv2D)             (None, 224, 224, 64) 36864       leaky_re_lu_196[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_197 (BatchN (None, 224, 224, 64) 256         conv2d_200[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_197 (LeakyReLU)     (None, 224, 224, 64) 0           batch_normalization_197[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_201 (Conv2D)             (None, 224, 224, 64) 36864       leaky_re_lu_197[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_68 (Add)                    (None, 224, 224, 64) 0           conv2d_201[0][0]                 \n",
      "                                                                 conv2d_199[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_198 (BatchN (None, 224, 224, 64) 256         add_68[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_198 (LeakyReLU)     (None, 224, 224, 64) 0           batch_normalization_198[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_202 (Conv2D)             (None, 224, 224, 64) 36864       leaky_re_lu_198[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_199 (BatchN (None, 224, 224, 64) 256         conv2d_202[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_199 (LeakyReLU)     (None, 224, 224, 64) 0           batch_normalization_199[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_203 (Conv2D)             (None, 224, 224, 64) 36864       leaky_re_lu_199[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_69 (Add)                    (None, 224, 224, 64) 0           conv2d_203[0][0]                 \n",
      "                                                                 add_68[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_200 (BatchN (None, 224, 224, 64) 256         add_69[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_200 (LeakyReLU)     (None, 224, 224, 64) 0           batch_normalization_200[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_204 (Conv2D)             (None, 112, 112, 128 73728       leaky_re_lu_200[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_201 (BatchN (None, 112, 112, 128 512         conv2d_204[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_201 (LeakyReLU)     (None, 112, 112, 128 0           batch_normalization_201[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_205 (Conv2D)             (None, 112, 112, 128 147456      leaky_re_lu_201[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_202 (BatchN (None, 112, 112, 128 512         conv2d_205[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_202 (LeakyReLU)     (None, 112, 112, 128 0           batch_normalization_202[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_206 (Conv2D)             (None, 112, 112, 128 147456      leaky_re_lu_202[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_70 (Add)                    (None, 112, 112, 128 0           conv2d_206[0][0]                 \n",
      "                                                                 conv2d_204[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_203 (BatchN (None, 112, 112, 128 512         add_70[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_203 (LeakyReLU)     (None, 112, 112, 128 0           batch_normalization_203[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_207 (Conv2D)             (None, 112, 112, 128 147456      leaky_re_lu_203[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_204 (BatchN (None, 112, 112, 128 512         conv2d_207[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_204 (LeakyReLU)     (None, 112, 112, 128 0           batch_normalization_204[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_208 (Conv2D)             (None, 112, 112, 128 147456      leaky_re_lu_204[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_71 (Add)                    (None, 112, 112, 128 0           conv2d_208[0][0]                 \n",
      "                                                                 add_70[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_205 (BatchN (None, 112, 112, 128 512         add_71[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_205 (LeakyReLU)     (None, 112, 112, 128 0           batch_normalization_205[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_209 (Conv2D)             (None, 56, 56, 256)  294912      leaky_re_lu_205[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_206 (BatchN (None, 56, 56, 256)  1024        conv2d_209[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_206 (LeakyReLU)     (None, 56, 56, 256)  0           batch_normalization_206[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_210 (Conv2D)             (None, 56, 56, 256)  589824      leaky_re_lu_206[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_207 (BatchN (None, 56, 56, 256)  1024        conv2d_210[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_207 (LeakyReLU)     (None, 56, 56, 256)  0           batch_normalization_207[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_211 (Conv2D)             (None, 56, 56, 256)  589824      leaky_re_lu_207[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_72 (Add)                    (None, 56, 56, 256)  0           conv2d_211[0][0]                 \n",
      "                                                                 conv2d_209[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_208 (BatchN (None, 56, 56, 256)  1024        add_72[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_208 (LeakyReLU)     (None, 56, 56, 256)  0           batch_normalization_208[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_212 (Conv2D)             (None, 56, 56, 256)  589824      leaky_re_lu_208[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_209 (BatchN (None, 56, 56, 256)  1024        conv2d_212[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_209 (LeakyReLU)     (None, 56, 56, 256)  0           batch_normalization_209[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_213 (Conv2D)             (None, 56, 56, 256)  589824      leaky_re_lu_209[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_73 (Add)                    (None, 56, 56, 256)  0           conv2d_213[0][0]                 \n",
      "                                                                 add_72[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_210 (BatchN (None, 56, 56, 256)  1024        add_73[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_210 (LeakyReLU)     (None, 56, 56, 256)  0           batch_normalization_210[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_214 (Conv2D)             (None, 28, 28, 512)  1179648     leaky_re_lu_210[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_211 (BatchN (None, 28, 28, 512)  2048        conv2d_214[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_211 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_211[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_215 (Conv2D)             (None, 28, 28, 512)  2359296     leaky_re_lu_211[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_212 (BatchN (None, 28, 28, 512)  2048        conv2d_215[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_212 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_212[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_216 (Conv2D)             (None, 28, 28, 512)  2359296     leaky_re_lu_212[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_74 (Add)                    (None, 28, 28, 512)  0           conv2d_216[0][0]                 \n",
      "                                                                 conv2d_214[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_213 (BatchN (None, 28, 28, 512)  2048        add_74[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_213 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_213[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_217 (Conv2D)             (None, 28, 28, 512)  2359296     leaky_re_lu_213[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_214 (BatchN (None, 28, 28, 512)  2048        conv2d_217[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_214 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_214[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_218 (Conv2D)             (None, 28, 28, 512)  2359296     leaky_re_lu_214[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_75 (Add)                    (None, 28, 28, 512)  0           conv2d_218[0][0]                 \n",
      "                                                                 add_74[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_215 (BatchN (None, 28, 28, 512)  2048        add_75[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_215 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_215[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "dilated_conv_1 (Conv2D)         (None, 28, 28, 512)  2359808     leaky_re_lu_215[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_216 (BatchN (None, 28, 28, 512)  2048        dilated_conv_1[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_216 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_216[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "dilated_conv_2 (Conv2D)         (None, 28, 28, 512)  2359808     leaky_re_lu_216[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "add_76 (Add)                    (None, 28, 28, 512)  0           dilated_conv_2[0][0]             \n",
      "                                                                 dilated_conv_1[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_217 (BatchN (None, 28, 28, 512)  2048        add_76[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_217 (LeakyReLU)     (None, 28, 28, 512)  0           batch_normalization_217[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_219 (Conv2D)             (None, 28, 28, 384)  196992      leaky_re_lu_217[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_218 (BatchN (None, 28, 28, 384)  1536        conv2d_219[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_218 (LeakyReLU)     (None, 28, 28, 384)  0           batch_normalization_218[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_2 (Conv2DTrans (None, 112, 112, 256 6291712     leaky_re_lu_218[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_219 (BatchN (None, 112, 112, 256 1024        conv2d_transpose_2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "leaky_re_lu_219 (LeakyReLU)     (None, 112, 112, 256 0           batch_normalization_219[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_220 (Conv2D)             (None, 112, 112, 1)  257         leaky_re_lu_219[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "up_sampling2d_8 (UpSampling2D)  (None, 448, 448, 1)  0           conv2d_220[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 25,337,633\n",
      "Trainable params: 25,324,513\n",
      "Non-trainable params: 13,120\n",
      "__________________________________________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# define iou or jaccard loss function\n",
    "def iou_loss(y_true, y_pred):\n",
    "    y_true = tf.reshape(y_true, [-1])\n",
    "    y_pred = tf.reshape(y_pred, [-1])\n",
    "    intersection = tf.reduce_sum(y_true * y_pred)\n",
    "    score = (intersection + 1.) / (tf.reduce_sum(y_true) + tf.reduce_sum(y_pred) - intersection + 1.)\n",
    "    return 1 - score\n",
    "\n",
    "# combine bce loss and iou loss\n",
    "def iou_bce_loss(y_true, y_pred):\n",
    "    return 0.25 * keras.losses.binary_crossentropy(y_true, y_pred) + 0.75 * iou_loss(y_true, y_pred)\n",
    "\n",
    "# mean iou as a metric\n",
    "def mean_iou(y_true, y_pred):\n",
    "    y_pred = tf.round(y_pred)\n",
    "    intersect = tf.reduce_sum(y_true * y_pred, axis=[1, 2, 3])\n",
    "    union = tf.reduce_sum(y_true, axis=[1, 2, 3]) + tf.reduce_sum(y_pred, axis=[1, 2, 3])\n",
    "    smooth = tf.ones(tf.shape(intersect))\n",
    "    return tf.reduce_mean((intersect + smooth) / (union - intersect + smooth))\n",
    "\n",
    "# create network and compiler\n",
    "model = create_network(input_size=IMAGE_SIZE, channels=32, n_blocks=2, depth=4)\n",
    "model.compile(optimizer='adam',\n",
    "              loss=iou_bce_loss,\n",
    "              metrics=['accuracy', mean_iou])\n",
    "\n",
    "# cosine learning rate annealing\n",
    "def cosine_annealing(x):\n",
    "    lr = 0.001\n",
    "    epochs = 20\n",
    "    return lr*(np.cos(np.pi*x/epochs)+1.)/2\n",
    "learning_rate = tf.keras.callbacks.LearningRateScheduler(cosine_annealing)\n",
    "checkpoint = keras.callbacks.ModelCheckpoint(CHECKPOINT_PATH, monitor='val_loss', verbose=1, save_best_only=False, save_weights_only=True, mode='auto', period=1)\n",
    "\n",
    "# create train and validation generators\n",
    "folder = './stage_1_train_images'\n",
    "train_gen = generator(folder, train_filenames, pneumonia_locations, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=True, augment=True, predict=False)\n",
    "valid_gen = generator(folder, valid_filenames, pneumonia_locations, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=False, predict=False)\n",
    "\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 576
    },
    "colab_type": "code",
    "id": "fM5VeiuDcxUY",
    "outputId": "3cb4027e-6753-45f0-d568-5bd4e611ea8d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.5060 - acc: 0.9610 - mean_iou: 0.5926\n",
      "Epoch 00001: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2564s 1s/step - loss: 0.5060 - acc: 0.9610 - mean_iou: 0.5926 - val_loss: 0.4852 - val_acc: 0.9669 - val_mean_iou: 0.6576\n",
      "Epoch 2/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.4758 - acc: 0.9634 - mean_iou: 0.6256\n",
      "Epoch 00002: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2537s 1s/step - loss: 0.4758 - acc: 0.9634 - mean_iou: 0.6255 - val_loss: 0.4651 - val_acc: 0.9647 - val_mean_iou: 0.6460\n",
      "Epoch 3/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.4635 - acc: 0.9649 - mean_iou: 0.6413\n",
      "Epoch 00003: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2537s 1s/step - loss: 0.4635 - acc: 0.9649 - mean_iou: 0.6413 - val_loss: 0.4620 - val_acc: 0.9700 - val_mean_iou: 0.6940\n",
      "Epoch 4/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.4531 - acc: 0.9664 - mean_iou: 0.6565\n",
      "Epoch 00004: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2537s 1s/step - loss: 0.4531 - acc: 0.9664 - mean_iou: 0.6565 - val_loss: 0.4504 - val_acc: 0.9646 - val_mean_iou: 0.6294\n",
      "Epoch 5/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.4412 - acc: 0.9679 - mean_iou: 0.6643\n",
      "Epoch 00005: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2536s 1s/step - loss: 0.4412 - acc: 0.9679 - mean_iou: 0.6644 - val_loss: 0.4432 - val_acc: 0.9739 - val_mean_iou: 0.7272\n",
      "Epoch 6/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.4342 - acc: 0.9684 - mean_iou: 0.6766\n",
      "Epoch 00006: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2537s 1s/step - loss: 0.4341 - acc: 0.9685 - mean_iou: 0.6767 - val_loss: 0.4351 - val_acc: 0.9726 - val_mean_iou: 0.7264\n",
      "Epoch 7/15\n",
      "1926/1927 [============================>.] - ETA: 1s - loss: 0.4296 - acc: 0.9696 - mean_iou: 0.6860\n",
      "Epoch 00007: saving model to model6.h5\n",
      "1927/1927 [==============================] - 2539s 1s/step - loss: 0.4295 - acc: 0.9695 - mean_iou: 0.6859 - val_loss: 0.4271 - val_acc: 0.9713 - val_mean_iou: 0.6860\n",
      "Epoch 8/15\n",
      " 281/1927 [===>..........................] - ETA: 34:17 - loss: 0.4276 - acc: 0.9686 - mean_iou: 0.6992"
     ]
    }
   ],
   "source": [
    "history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate, checkpoint], epochs=15, shuffle=True, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1106
    },
    "colab_type": "code",
    "id": "7tyG7P79Y42U",
    "outputId": "b5bbde67-09f1-4d4e-da3a-4bb4b397a9a8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "1926/1927 [============================>.] - ETA: 0s - loss: 0.4135 - acc: 0.9709 - mean_iou: 0.7030\n",
      "Epoch 00001: saving model to model5.h5\n",
      "1927/1927 [==============================] - 1895s 983ms/step - loss: 0.4135 - acc: 0.9709 - mean_iou: 0.7030 - val_loss: 0.4153 - val_acc: 0.9727 - val_mean_iou: 0.6989\n",
      "Epoch 2/5\n",
      "1926/1927 [============================>.] - ETA: 0s - loss: 0.4129 - acc: 0.9713 - mean_iou: 0.7080\n",
      "Epoch 00002: saving model to model5.h5\n",
      "1927/1927 [==============================] - 1896s 984ms/step - loss: 0.4129 - acc: 0.9713 - mean_iou: 0.7080 - val_loss: 0.4190 - val_acc: 0.9684 - val_mean_iou: 0.6962\n",
      "Epoch 3/5\n",
      "1926/1927 [============================>.] - ETA: 0s - loss: 0.4114 - acc: 0.9711 - mean_iou: 0.7029\n",
      "Epoch 00003: saving model to model5.h5\n",
      "1927/1927 [==============================] - 1896s 984ms/step - loss: 0.4115 - acc: 0.9711 - mean_iou: 0.7030 - val_loss: 0.4173 - val_acc: 0.9741 - val_mean_iou: 0.7276\n",
      "Epoch 4/5\n",
      "1808/1927 [===========================>..] - ETA: 1:49 - loss: 0.4090 - acc: 0.9710 - mean_iou: 0.7056"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "ignored",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-30-b105edd4ccb8>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mhistory\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit_generator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_gen\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalidation_data\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mvalid_gen\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcallbacks\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlearning_rate\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcheckpoint\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mepochs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(self, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m   1777\u001b[0m         \u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1778\u001b[0m         \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mshuffle\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1779\u001b[0;31m         initial_epoch=initial_epoch)\n\u001b[0m\u001b[1;32m   1780\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1781\u001b[0m   def evaluate_generator(self,\n",
      "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training_generator.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(model, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m    202\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    203\u001b[0m         outs = model.train_on_batch(\n\u001b[0;32m--> 204\u001b[0;31m             x, y, sample_weight=sample_weight, class_weight=class_weight)\n\u001b[0m\u001b[1;32m    205\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    206\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mouts\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py\u001b[0m in \u001b[0;36mtrain_on_batch\u001b[0;34m(self, x, y, sample_weight, class_weight)\u001b[0m\n\u001b[1;32m   1550\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1551\u001b[0m       \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_train_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1552\u001b[0;31m       \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mins\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1553\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1554\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/backend.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m   2912\u001b[0m       \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_callable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfeed_arrays\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_symbols\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msymbol_vals\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msession\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2913\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2914\u001b[0;31m     \u001b[0mfetched\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_callable_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0marray_vals\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2915\u001b[0m     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call_fetch_callbacks\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfetched\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_fetches\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2916\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mfetched\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1380\u001b[0m           ret = tf_session.TF_SessionRunCallable(\n\u001b[1;32m   1381\u001b[0m               \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_handle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstatus\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1382\u001b[0;31m               run_metadata_ptr)\n\u001b[0m\u001b[1;32m   1383\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1384\u001b[0m           \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "# history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate, checkpoint], epochs=5, shuffle=True, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 265
    },
    "colab_type": "code",
    "id": "y_bLpJPBcxUh",
    "outputId": "114b8b85-1a77-4c3f-965a-6adedd295043"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsYAAAD4CAYAAADvhyBBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9a60652cc0>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,4))\n",
    "plt.subplot(131)\n",
    "plt.plot(history.epoch, history.history[\"loss\"], label=\"Train loss\")\n",
    "plt.plot(history.epoch, history.history[\"val_loss\"], label=\"Valid loss\")\n",
    "plt.legend()\n",
    "plt.subplot(132)\n",
    "plt.plot(history.epoch, history.history[\"acc\"], label=\"Train accuracy\")\n",
    "plt.plot(history.epoch, history.history[\"val_acc\"], label=\"Valid accuracy\")\n",
    "plt.legend()\n",
    "plt.subplot(133)\n",
    "plt.plot(history.epoch, history.history[\"mean_iou\"], label=\"Train iou\")\n",
    "plt.plot(history.epoch, history.history[\"val_mean_iou\"], label=\"Valid iou\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3tAntxescxUq"
   },
   "source": [
    "# Predict test images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 36
    },
    "colab_type": "code",
    "id": "fnu7ebfScxUu",
    "outputId": "03b263f4-c6ca-4cb7-c15a-802d1b1829dc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n test samples: 1000\n"
     ]
    }
   ],
   "source": [
    "# load and shuffle filenames\n",
    "folder = './stage_1_test_images'\n",
    "test_filenames = os.listdir(folder)\n",
    "print('n test samples:', len(test_filenames))\n",
    "\n",
    "# create test generator with predict flag set to True\n",
    "test_gen = generator(folder, test_filenames, None, batch_size=25, image_size=IMAGE_SIZE, shuffle=False, predict=True)\n",
    "\n",
    "# create submission dictionary\n",
    "submission_dict = {}\n",
    "# loop through testset\n",
    "for imgs, filenames in test_gen:\n",
    "    # predict batch of images\n",
    "    preds = model.predict(imgs)\n",
    "    # loop through batch\n",
    "    for pred, filename in zip(preds, filenames):\n",
    "        # resize predicted mask\n",
    "        pred = resize(pred, (1024, 1024), mode='reflect')\n",
    "        # threshold predicted mask\n",
    "        comp = pred[:, :, 0] > 0.5\n",
    "        # apply connected components\n",
    "        comp = measure.label(comp)\n",
    "        # apply bounding boxes\n",
    "        predictionString = ''\n",
    "        for region in measure.regionprops(comp):\n",
    "            # retrieve x, y, height and width\n",
    "            y, x, y2, x2 = region.bbox\n",
    "            height = y2 - y\n",
    "            width = x2 - x\n",
    "            # proxy for confidence score\n",
    "            conf = np.mean(pred[y:y+height, x:x+width])\n",
    "            # add to predictionString\n",
    "            predictionString += str(conf) + ' ' + str(x) + ' ' + str(y) + ' ' + str(width) + ' ' + str(height) + ' '\n",
    "        # add filename and predictionString to dictionary\n",
    "        filename = filename.split('.')[0]\n",
    "        submission_dict[filename] = predictionString\n",
    "    # stop if we've got them all\n",
    "    if len(submission_dict) >= len(test_filenames):\n",
    "        break\n",
    "\n",
    "# save dictionary as csv file\n",
    "sub = pd.DataFrame.from_dict(submission_dict,orient='index')\n",
    "sub.index.names = ['patientId']\n",
    "sub.columns = ['PredictionString']\n",
    "sub.to_csv('submission.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 36
    },
    "colab_type": "code",
    "id": "U1NEg9QCcxU4",
    "outputId": "1ee5405d-aff4-4b46-8886-fdc61f96c103"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully submitted to RSNA Pneumonia Detection Challenge"
     ]
    }
   ],
   "source": [
    "!kaggle competitions submit -c rsna-pneumonia-detection-challenge -f submission.csv -m \"Colab segmentation 384x384 20 epochs\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 73
    },
    "colab_type": "code",
    "id": "NTQ_vDaVfvZR",
    "outputId": "b48a94f3-af86-4ea2-a88a-f36fa6fe09ed"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File ID: 1p6FODG73nheBG4yxeipGDHKVrPLlelK7\n",
      "File ID: 14GakqSIHa1ypJ4KA5B_uHmysxTr2Dxit\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'id': '14GakqSIHa1ypJ4KA5B_uHmysxTr2Dxit'}"
      ]
     },
     "execution_count": 33,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "save_file_to_drive(\"submission.csv\", \"submission.csv\")\n",
    "save_file_to_drive(CHECKPOINT_PATH, CHECKPOINT_PATH)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 36
    },
    "colab_type": "code",
    "id": "yf06y04TVRi2",
    "outputId": "2e4bb87f-f402-4bb5-a162-2be7e2088997"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Updated property [core/project].\r\n"
     ]
    }
   ],
   "source": [
    "# upload checkpoint to GCS\n",
    "project_id = 'mammography-198911'\n",
    "bucket_name = 'pneumonia'\n",
    "\n",
    "!gcloud config set project {project_id}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 73
    },
    "colab_type": "code",
    "id": "acm5YWVLVSQD",
    "outputId": "24fa6787-fafd-477f-9d6f-be3924f57f26"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Copying file://./model5.h5 [Content-Type=application/octet-stream]...\n",
      "\\\n",
      "Operation completed over 1 objects/66.8 MiB.                                     \n"
     ]
    }
   ],
   "source": [
    "!gsutil cp ./{CHECKPOINT_PATH} gs://{bucket_name}/"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "best-cnn-segmentation-connected-components-384x384.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}