cnn_segmentation_connected_components_4.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "cnn-segmentation-connected-components.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "sliy7m7L-M-D",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "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 nodules.\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."
      ]
    },
    {
      "metadata": {
        "id": "xjzfSdjV-Y3y",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# install dependencies not included by Colab\n",
        "# use pip3 to ensure compatibility w/ Google Deep Learning Images \n",
        "!pip3 install -q pydicom \n",
        "!pip3 install -q tqdm \n",
        "!pip3 install -q imgaug \n",
        "!pip3 install -q kaggle"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "pu1pks8K-M-G",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "ea06f512-c642-40af-f5ee-fb5ebf0c88a6"
      },
      "cell_type": "code",
      "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",
        "from keras import layers\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "from matplotlib import pyplot as plt"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "ZFNwjWWc_JgL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "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"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ejrE2hQw-fMG",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# enter your Kaggle credentionals here\n",
        "os.environ['KAGGLE_USERNAME']=\"skooch\"\n",
        "os.environ['KAGGLE_KEY']=\"42f8a02ee92cc773d1dbe66565673ad3\""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "BJskmI-Y-fwY",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Root directory of the project\n",
        "ROOT_DIR = os.path.abspath('./data')\n",
        "\n",
        "# Directory to save logs and trained model\n",
        "MODEL_DIR = os.path.join(ROOT_DIR, 'logs')\n",
        "\n",
        "if not os.path.exists(ROOT_DIR):\n",
        "    os.makedirs(ROOT_DIR)\n",
        "os.chdir(ROOT_DIR)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "-tBhkYuWKWic",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "!ls"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "NonqWakT-jTC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        },
        "outputId": "08fa2aa4-d90b-410b-f543-884ba7ddd125"
      },
      "cell_type": "code",
      "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",
        "!kaggle competitions download -c rsna-pneumonia-detection-challenge"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading GCP%20Credits%20Request%20Link%20-%20RSNA.txt to /content/data\r\n",
            "\r  0%|                                                | 0.00/55.0 [00:00<?, ?B/s]\r\n",
            "100%|████████████████████████████████████████| 55.0/55.0 [00:00<00:00, 35.5kB/s]\n",
            "Downloading stage_1_detailed_class_info.csv.zip to /content/data\n",
            "  0%|                                                | 0.00/561k [00:00<?, ?B/s]\n",
            "100%|████████████████████████████████████████| 561k/561k [00:00<00:00, 63.5MB/s]\n",
            "Downloading stage_1_sample_submission.csv to /content/data\n",
            "  0%|                                               | 0.00/51.8k [00:00<?, ?B/s]\n",
            "100%|██████████████████████████████████████| 51.8k/51.8k [00:00<00:00, 34.2MB/s]\n",
            "Downloading stage_1_train_labels.csv.zip to /content/data\n",
            "  0%|                                                | 0.00/634k [00:00<?, ?B/s]\n",
            "100%|████████████████████████████████████████| 634k/634k [00:00<00:00, 94.2MB/s]\n",
            "Downloading stage_1_test_images.zip to /content/data\n",
            " 93%|██████████████████████████████████████▎  | 118M/126M [00:00<00:00, 126MB/s]\n",
            "100%|█████████████████████████████████████████| 126M/126M [00:00<00:00, 153MB/s]\n",
            "Downloading stage_1_train_images.zip to /content/data\n",
            "100%|██████████████████████████████████████▉| 3.17G/3.17G [00:24<00:00, 150MB/s]\n",
            "100%|███████████████████████████████████████| 3.17G/3.17G [00:24<00:00, 140MB/s]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "YHvOAY1M-pzE",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# unzipping takes a few minutes\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"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "EQG4ghkH-M-e",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Load nodule locations\n",
        "\n",
        "Table contains [filename : nodule location] pairs per row. \n",
        "* If a filename contains multiple nodules, the table contains multiple rows with the same filename but different nodule locations. \n",
        "* If a filename contains no nodules it contains a single row with an empty nodule 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 nodule locations in that filename as value. \n",
        "* If a filename is not present in the dictionary it means that it contains no nodules."
      ]
    },
    {
      "metadata": {
        "id": "67McqwYt-t5W",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "e2fdbd9e-870b-4db3-f089-4e753885f2cc"
      },
      "cell_type": "code",
      "source": [
        "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)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train dir: /content/data/stage_1_train_images\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "HIYblxy8-M-f",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# empty dictionary\n",
        "nodule_locations = {}\n",
        "# load table\n",
        "with open('./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",
        "        nodule = rows[5]\n",
        "        # if row contains a nodule add label to dictionary\n",
        "        # which contains a list of nodule locations per filename\n",
        "        if nodule == '1':\n",
        "            # convert string to float to int\n",
        "            location = [int(float(i)) for i in location]\n",
        "            # save nodule location in dictionary\n",
        "            if filename in nodule_locations:\n",
        "                nodule_locations[filename].append(location)\n",
        "            else:\n",
        "                nodule_locations[filename] = [location]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "guL6FY9N-M-m",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Load filenames"
      ]
    },
    {
      "metadata": {
        "id": "ebHtuAkP-M-s",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "cfe57e10-c83a-4270-a45c-f5be00544635"
      },
      "cell_type": "code",
      "source": [
        "# load and shuffle filenames\n",
        "folder = train_dicom_dir\n",
        "filenames = os.listdir(folder)\n",
        "random.shuffle(filenames)\n",
        "# split into train and validation filenames\n",
        "n_valid_samples = 2560\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"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "n train samples 23124\n",
            "n valid samples 2560\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "XRTP_RT6-M-7",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "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",
        "    "
      ]
    },
    {
      "metadata": {
        "id": "mNHOs4vXGdwJ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 12\n",
        "IMAGE_SIZE = 384\n",
        "CHECKPOINT_PATH = \"model2_4.h5\""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "aD51VT0I-M-9",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class generator(keras.utils.Sequence):\n",
        "    \n",
        "    def __init__(self, folder, filenames, nodule_locations=None, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=True, predict=False):\n",
        "        self.folder = folder\n",
        "        self.filenames = filenames\n",
        "        self.nodule_locations = nodule_locations\n",
        "        self.batch_size = batch_size\n",
        "        self.image_size = image_size\n",
        "        self.shuffle = shuffle\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 nodules\n",
        "        if filename in nodule_locations:\n",
        "            # loop through nodules\n",
        "            for location in nodule_locations[filename]:\n",
        "                # add 1's at the location of the nodule\n",
        "                x, y, w, h = location\n",
        "                msk[y:y+h, x:x+w] = 1\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)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "UjoLi1xYiOEp",
        "colab_type": "code",
        "colab": {
          "resources": {
            "http://localhost:8080/nbextensions/google.colab/files.js": {
              "data": "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",
              "ok": true,
              "headers": [
                [
                  "content-type",
                  "application/javascript"
                ]
              ],
              "status": 200,
              "status_text": ""
            }
          },
          "base_uri": "https://localhost:8080/",
          "height": 72
        },
        "outputId": "81bc882e-6376-46b7-c05d-2741b91ca879"
      },
      "cell_type": "code",
      "source": [
        "# upload last checkpoint\n",
        "# from google.colab import files\n",
        "\n",
        "# uploaded = files.upload()"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-d1040cec-c7d9-4c4e-a74e-b102d4867733\" name=\"files[]\" multiple disabled />\n",
              "     <output id=\"result-d1040cec-c7d9-4c4e-a74e-b102d4867733\">\n",
              "      Upload widget is only available when the cell has been executed in the\n",
              "      current browser session. Please rerun this cell to enable.\n",
              "      </output>\n",
              "      <script src=\"/nbextensions/google.colab/files.js\"></script> "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Saving model2_3.h5 to model2_3 (1).h5\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "RbHHFLzb-M_F",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Network"
      ]
    },
    {
      "metadata": {
        "id": "rYHu1v6V-M_R",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "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",
        "    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",
        "    # residual blocks\n",
        "    for d in range(depth):\n",
        "        channels = channels * 2\n",
        "        if channels > 512:\n",
        "          channels = 512\n",
        "        \n",
        "        x = create_downsample(channels, x)\n",
        "        for b in range(n_blocks):\n",
        "            x = create_resblock(channels, x)\n",
        "    # output\n",
        "    x = keras.layers.BatchNormalization(momentum=0.9)(x)\n",
        "    x = keras.layers.LeakyReLU(0)(x)\n",
        "    \n",
        "    # downsample one last time\n",
        "#     x = keras.layers.Conv2D(512, (3,3), strides=(2,2), padding='same', activation='relu')(x)\n",
        "    \n",
        "    # reduce number of feature maps\n",
        "    x = keras.layers.Conv2D(384, (3,3), padding='same', activation='relu')(x)\n",
        "\n",
        "    #     x = keras.layers.Dropout(0.125)(x)\n",
        "    \n",
        "    # deconv to 4x size\n",
        "    x = keras.layers.Conv2DTranspose(64, (8,8), (4,4), padding=\"same\", activation=\"relu\")(x)\n",
        "    \n",
        "    # upsample\n",
        "    x = keras.layers.UpSampling2D(2**(depth - 1))(x)\n",
        "    \n",
        "    # 1 transpose conv to get back to original size\n",
        "    outputs = keras.layers.Conv2D(1, 1, activation='sigmoid')(x)\n",
        "    model = keras.Model(inputs=inputs, outputs=outputs)\n",
        "    return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "wsBfvg2Y-M_e",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Train network\n"
      ]
    },
    {
      "metadata": {
        "id": "qkxSFJ9w-M_f",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 3043
        },
        "outputId": "614bef22-b3ce-4ad6-efad-9681c625e2aa"
      },
      "cell_type": "code",
      "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 0.3 * keras.losses.binary_crossentropy(y_true, y_pred) + 0.7 * 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=24, n_blocks=2, depth=4)\n",
        "print(model.summary())\n",
        "\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 = 0.003\n",
        "    epochs = 25\n",
        "    return lr*(np.cos(np.pi*x/epochs)+1.)/2\n",
        "  \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=False, mode='auto', period=1)\n",
        "\n",
        "\n",
        "# create train and validation generators\n",
        "folder = train_dicom_dir\n",
        "train_gen = generator(folder, train_filenames, nodule_locations, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=True, predict=False)\n",
        "valid_gen = generator(folder, valid_filenames, nodule_locations, batch_size=BATCH_SIZE, image_size=IMAGE_SIZE, shuffle=False, predict=False)\n"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            (None, 384, 384, 1)  0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d (Conv2D)                 (None, 384, 384, 24) 216         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization (BatchNorma (None, 384, 384, 24) 96          conv2d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu (LeakyReLU)         (None, 384, 384, 24) 0           batch_normalization[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 384, 384, 48) 1152        leaky_re_lu[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D)    (None, 192, 192, 48) 0           conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_1 (BatchNor (None, 192, 192, 48) 192         max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_1 (LeakyReLU)       (None, 192, 192, 48) 0           batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 192, 192, 48) 20736       leaky_re_lu_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_2 (BatchNor (None, 192, 192, 48) 192         conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_2 (LeakyReLU)       (None, 192, 192, 48) 0           batch_normalization_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 192, 192, 48) 20736       leaky_re_lu_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "add (Add)                       (None, 192, 192, 48) 0           conv2d_3[0][0]                   \n",
            "                                                                 max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_3 (BatchNor (None, 192, 192, 48) 192         add[0][0]                        \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_3 (LeakyReLU)       (None, 192, 192, 48) 0           batch_normalization_3[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_4 (Conv2D)               (None, 192, 192, 48) 20736       leaky_re_lu_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_4 (BatchNor (None, 192, 192, 48) 192         conv2d_4[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_4 (LeakyReLU)       (None, 192, 192, 48) 0           batch_normalization_4[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_5 (Conv2D)               (None, 192, 192, 48) 20736       leaky_re_lu_4[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "add_1 (Add)                     (None, 192, 192, 48) 0           conv2d_5[0][0]                   \n",
            "                                                                 add[0][0]                        \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_5 (BatchNor (None, 192, 192, 48) 192         add_1[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_5 (LeakyReLU)       (None, 192, 192, 48) 0           batch_normalization_5[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_6 (Conv2D)               (None, 192, 192, 96) 4608        leaky_re_lu_5[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2D)  (None, 96, 96, 96)   0           conv2d_6[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_6 (BatchNor (None, 96, 96, 96)   384         max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_6 (LeakyReLU)       (None, 96, 96, 96)   0           batch_normalization_6[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_7 (Conv2D)               (None, 96, 96, 96)   82944       leaky_re_lu_6[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_7 (BatchNor (None, 96, 96, 96)   384         conv2d_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_7 (LeakyReLU)       (None, 96, 96, 96)   0           batch_normalization_7[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_8 (Conv2D)               (None, 96, 96, 96)   82944       leaky_re_lu_7[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "add_2 (Add)                     (None, 96, 96, 96)   0           conv2d_8[0][0]                   \n",
            "                                                                 max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_8 (BatchNor (None, 96, 96, 96)   384         add_2[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_8 (LeakyReLU)       (None, 96, 96, 96)   0           batch_normalization_8[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_9 (Conv2D)               (None, 96, 96, 96)   82944       leaky_re_lu_8[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_9 (BatchNor (None, 96, 96, 96)   384         conv2d_9[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_9 (LeakyReLU)       (None, 96, 96, 96)   0           batch_normalization_9[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_10 (Conv2D)              (None, 96, 96, 96)   82944       leaky_re_lu_9[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "add_3 (Add)                     (None, 96, 96, 96)   0           conv2d_10[0][0]                  \n",
            "                                                                 add_2[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_10 (BatchNo (None, 96, 96, 96)   384         add_3[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_10 (LeakyReLU)      (None, 96, 96, 96)   0           batch_normalization_10[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_11 (Conv2D)              (None, 96, 96, 192)  18432       leaky_re_lu_10[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2D)  (None, 48, 48, 192)  0           conv2d_11[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_11 (BatchNo (None, 48, 48, 192)  768         max_pooling2d_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_11 (LeakyReLU)      (None, 48, 48, 192)  0           batch_normalization_11[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_12 (Conv2D)              (None, 48, 48, 192)  331776      leaky_re_lu_11[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_12 (BatchNo (None, 48, 48, 192)  768         conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_12 (LeakyReLU)      (None, 48, 48, 192)  0           batch_normalization_12[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_13 (Conv2D)              (None, 48, 48, 192)  331776      leaky_re_lu_12[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "add_4 (Add)                     (None, 48, 48, 192)  0           conv2d_13[0][0]                  \n",
            "                                                                 max_pooling2d_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_13 (BatchNo (None, 48, 48, 192)  768         add_4[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_13 (LeakyReLU)      (None, 48, 48, 192)  0           batch_normalization_13[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_14 (Conv2D)              (None, 48, 48, 192)  331776      leaky_re_lu_13[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_14 (BatchNo (None, 48, 48, 192)  768         conv2d_14[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_14 (LeakyReLU)      (None, 48, 48, 192)  0           batch_normalization_14[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_15 (Conv2D)              (None, 48, 48, 192)  331776      leaky_re_lu_14[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "add_5 (Add)                     (None, 48, 48, 192)  0           conv2d_15[0][0]                  \n",
            "                                                                 add_4[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_15 (BatchNo (None, 48, 48, 192)  768         add_5[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_15 (LeakyReLU)      (None, 48, 48, 192)  0           batch_normalization_15[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_16 (Conv2D)              (None, 48, 48, 384)  73728       leaky_re_lu_15[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2D)  (None, 24, 24, 384)  0           conv2d_16[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_16 (BatchNo (None, 24, 24, 384)  1536        max_pooling2d_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_16 (LeakyReLU)      (None, 24, 24, 384)  0           batch_normalization_16[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_17 (Conv2D)              (None, 24, 24, 384)  1327104     leaky_re_lu_16[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_17 (BatchNo (None, 24, 24, 384)  1536        conv2d_17[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_17 (LeakyReLU)      (None, 24, 24, 384)  0           batch_normalization_17[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_18 (Conv2D)              (None, 24, 24, 384)  1327104     leaky_re_lu_17[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "add_6 (Add)                     (None, 24, 24, 384)  0           conv2d_18[0][0]                  \n",
            "                                                                 max_pooling2d_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_18 (BatchNo (None, 24, 24, 384)  1536        add_6[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_18 (LeakyReLU)      (None, 24, 24, 384)  0           batch_normalization_18[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_19 (Conv2D)              (None, 24, 24, 384)  1327104     leaky_re_lu_18[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_19 (BatchNo (None, 24, 24, 384)  1536        conv2d_19[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_19 (LeakyReLU)      (None, 24, 24, 384)  0           batch_normalization_19[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_20 (Conv2D)              (None, 24, 24, 384)  1327104     leaky_re_lu_19[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "add_7 (Add)                     (None, 24, 24, 384)  0           conv2d_20[0][0]                  \n",
            "                                                                 add_6[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_20 (BatchNo (None, 24, 24, 384)  1536        add_7[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "leaky_re_lu_20 (LeakyReLU)      (None, 24, 24, 384)  0           batch_normalization_20[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_21 (Conv2D)              (None, 12, 12, 512)  1769984     leaky_re_lu_20[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_22 (Conv2D)              (None, 12, 12, 256)  1179904     conv2d_21[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose (Conv2DTranspo (None, 48, 48, 128)  2097280     conv2d_22[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "up_sampling2d (UpSampling2D)    (None, 384, 384, 128 0           conv2d_transpose[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_23 (Conv2D)              (None, 384, 384, 1)  129         up_sampling2d[0][0]              \n",
            "==================================================================================================\n",
            "Total params: 12,210,169\n",
            "Trainable params: 12,202,921\n",
            "Non-trainable params: 7,248\n",
            "__________________________________________________________________________________________________\n",
            "None\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "KKSw1v5GE5bX",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "try:\n",
        "  model.load_weights(CHECKPOINT_PATH)\n",
        "except:\n",
        "  pass\n",
        "# model.load_weights(CHECKPOINT_PATH)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "A3spdZG5cBsF",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 697
        },
        "outputId": "86c03c40-0176-4674-a200-639d4986b324"
      },
      "cell_type": "code",
      "source": [
        "history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate, checkpoint], epochs=10, shuffle=True, verbose=1)"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.4102 - acc: 0.9717 - mean_iou: 0.6953\n",
            "Epoch 00001: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2545s 1s/step - loss: 0.4103 - acc: 0.9717 - mean_iou: 0.6953 - val_loss: 0.4197 - val_acc: 0.9767 - val_mean_iou: 0.7572\n",
            "Epoch 2/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.4006 - acc: 0.9727 - mean_iou: 0.7081\n",
            "Epoch 00002: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2505s 1s/step - loss: 0.4006 - acc: 0.9727 - mean_iou: 0.7081 - val_loss: 0.3989 - val_acc: 0.9746 - val_mean_iou: 0.7277\n",
            "Epoch 3/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3977 - acc: 0.9731 - mean_iou: 0.7102\n",
            "Epoch 00003: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2517s 1s/step - loss: 0.3976 - acc: 0.9731 - mean_iou: 0.7102 - val_loss: 0.4050 - val_acc: 0.9692 - val_mean_iou: 0.6693\n",
            "Epoch 4/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3984 - acc: 0.9733 - mean_iou: 0.7082\n",
            "Epoch 00004: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2543s 1s/step - loss: 0.3984 - acc: 0.9733 - mean_iou: 0.7080 - val_loss: 0.4103 - val_acc: 0.9666 - val_mean_iou: 0.6750\n",
            "Epoch 5/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3936 - acc: 0.9736 - mean_iou: 0.7126\n",
            "Epoch 00005: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2524s 1s/step - loss: 0.3936 - acc: 0.9736 - mean_iou: 0.7126 - val_loss: 0.3956 - val_acc: 0.9746 - val_mean_iou: 0.7385\n",
            "Epoch 6/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3899 - acc: 0.9741 - mean_iou: 0.7166\n",
            "Epoch 00006: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2513s 1s/step - loss: 0.3898 - acc: 0.9741 - mean_iou: 0.7167 - val_loss: 0.3960 - val_acc: 0.9701 - val_mean_iou: 0.6602\n",
            "Epoch 7/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3883 - acc: 0.9745 - mean_iou: 0.7194\n",
            "Epoch 00007: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2521s 1s/step - loss: 0.3884 - acc: 0.9745 - mean_iou: 0.7195 - val_loss: 0.3912 - val_acc: 0.9738 - val_mean_iou: 0.7153\n",
            "Epoch 8/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3848 - acc: 0.9749 - mean_iou: 0.7250\n",
            "Epoch 00008: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2502s 1s/step - loss: 0.3849 - acc: 0.9749 - mean_iou: 0.7251 - val_loss: 0.3950 - val_acc: 0.9751 - val_mean_iou: 0.7389\n",
            "Epoch 9/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3815 - acc: 0.9753 - mean_iou: 0.7272\n",
            "Epoch 00009: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2523s 1s/step - loss: 0.3815 - acc: 0.9753 - mean_iou: 0.7272 - val_loss: 0.3928 - val_acc: 0.9729 - val_mean_iou: 0.7125\n",
            "Epoch 10/10\n",
            "1926/1927 [============================>.] - ETA: 1s - loss: 0.3784 - acc: 0.9760 - mean_iou: 0.7348\n",
            "Epoch 00010: saving model to model2.h5\n",
            "1927/1927 [==============================] - 2515s 1s/step - loss: 0.3784 - acc: 0.9760 - mean_iou: 0.7349 - val_loss: 0.4116 - val_acc: 0.9773 - val_mean_iou: 0.7592\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "5sg5Qj3JW1QI",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "59397761-1238-44d0-f74b-f647b109ed58"
      },
      "cell_type": "code",
      "source": [
        "history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate, checkpoint], epochs=2, shuffle=True, verbose=1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/2\n",
            " 493/1927 [======>.......................] - ETA: 29:17 - loss: 0.3761 - acc: 0.9754 - mean_iou: 0.7348"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "CO9WGS29Nbnu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "02673e37-143e-4867-84a1-889e53f07898"
      },
      "cell_type": "code",
      "source": [
        "# history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate, checkpoint], epochs=5, shuffle=True, verbose=1)"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.4055 - acc: 0.9719 - mean_iou: 0.7040\n",
            "Epoch 00001: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1869s 970ms/step - loss: 0.4055 - acc: 0.9719 - mean_iou: 0.7041 - val_loss: 0.4192 - val_acc: 0.9712 - val_mean_iou: 0.6853\n",
            "Epoch 2/5\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.4045 - acc: 0.9722 - mean_iou: 0.6996\n",
            "Epoch 00002: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1868s 969ms/step - loss: 0.4045 - acc: 0.9722 - mean_iou: 0.6996 - val_loss: 0.4195 - val_acc: 0.9711 - val_mean_iou: 0.6862\n",
            "Epoch 3/5\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.4027 - acc: 0.9728 - mean_iou: 0.7057\n",
            "Epoch 00003: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1867s 969ms/step - loss: 0.4028 - acc: 0.9728 - mean_iou: 0.7057 - val_loss: 0.4438 - val_acc: 0.9747 - val_mean_iou: 0.7335\n",
            "Epoch 4/5\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.4023 - acc: 0.9724 - mean_iou: 0.7059\n",
            "Epoch 00004: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1868s 969ms/step - loss: 0.4023 - acc: 0.9724 - mean_iou: 0.7059 - val_loss: 0.4492 - val_acc: 0.9749 - val_mean_iou: 0.7297\n",
            "Epoch 5/5\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.3977 - acc: 0.9732 - mean_iou: 0.7151\n",
            "Epoch 00005: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1866s 968ms/step - loss: 0.3977 - acc: 0.9732 - mean_iou: 0.7151 - val_loss: 0.4118 - val_acc: 0.9717 - val_mean_iou: 0.6975\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "y0oa3cNPcNI5",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "18d7d7bd-db18-4812-cc5d-65137680c825"
      },
      "cell_type": "code",
      "source": [
        "# history = model.fit_generator(train_gen, validation_data=valid_gen, callbacks=[learning_rate, checkpoint], epochs=3, shuffle=True, verbose=1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/3\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.3982 - acc: 0.9732 - mean_iou: 0.7083\n",
            "Epoch 00001: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1866s 969ms/step - loss: 0.3982 - acc: 0.9732 - mean_iou: 0.7083 - val_loss: 0.4057 - val_acc: 0.9713 - val_mean_iou: 0.7013\n",
            "Epoch 2/3\n",
            "1926/1927 [============================>.] - ETA: 0s - loss: 0.3979 - acc: 0.9733 - mean_iou: 0.7107\n",
            "Epoch 00002: saving model to model2.h5\n",
            "1927/1927 [==============================] - 1866s 968ms/step - loss: 0.3980 - acc: 0.9733 - mean_iou: 0.7107 - val_loss: 0.4131 - val_acc: 0.9723 - val_mean_iou: 0.6899\n",
            "Epoch 3/3\n",
            " 295/1927 [===>..........................] - ETA: 24:40 - loss: 0.3823 - acc: 0.9755 - mean_iou: 0.7298"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "yMDSkSHm-M_r",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "outputId": "ad4c4266-197b-48ad-ef55-fc079e686840"
      },
      "cell_type": "code",
      "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()"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsYAAAD4CAYAAADvhyBBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f0e98049dd8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "dz9at3ZH-NAE",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Predict test images"
      ]
    },
    {
      "metadata": {
        "id": "cCMcFaVN-NAF",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "1f9ca59e-7d3c-4ced-d0a1-3795d5c769d0"
      },
      "cell_type": "code",
      "source": [
        "# load and shuffle filenames\n",
        "folder = test_dicom_dir\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')"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "n test samples: 1000\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "whgCcttS_SUa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "eee71d29-2fa3-4e6a-b103-32039c455e7c"
      },
      "cell_type": "code",
      "source": [
        "!kaggle competitions submit -c rsna-pneumonia-detection-challenge -f submission.csv -m \"Segmentation from colab size 384 20 epochs\""
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Successfully submitted to RSNA Pneumonia Detection Challenge"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "D0UDLKjA_TZ7",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "8f857e6b-efe2-4f8b-b05f-41855eb7c8f9"
      },
      "cell_type": "code",
      "source": [
        "save_file_to_drive(\"submission.csv\", \"submission.csv\")"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "File ID: 1IELLnM94WsiRvdV2yz-F89Sy5usgbFL2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'id': '1IELLnM94WsiRvdV2yz-F89Sy5usgbFL2'}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        }
      ]
    },
    {
      "metadata": {
        "id": "20Yb5s_0yRUS",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "9322ecde-0583-4b5f-ac74-1c4f28807522"
      },
      "cell_type": "code",
      "source": [
        "save_file_to_drive(CHECKPOINT_PATH, CHECKPOINT_PATH)"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "File ID: 1oxs1xIp0LUEls8LOA5knii7KKU2Xj8XB\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'id': '1oxs1xIp0LUEls8LOA5knii7KKU2Xj8XB'}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "metadata": {
        "id": "JUFKH6Pjn4WJ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# upload checkpoint to GCS\n",
        "project_id = 'mammography-198911'\n",
        "bucket_name = 'pneumonia'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "fbXxZ-zOoNpE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "b83b4709-5be5-4783-ff9b-5c33ad61db50"
      },
      "cell_type": "code",
      "source": [
        "!gcloud config set project {project_id}\n",
        "!gsutil cp ./{CHECKPOINT_PATH} gs://{bucket_name}/"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Updated property [core/project].\n",
            "Copying file://./model2_3.h5 [Content-Type=application/octet-stream]...\n",
            "-\n",
            "Operation completed over 1 objects/2.4 MiB.                                      \n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}