Machine_Learning/code_examples/unet_segmentation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "unet-segmentation.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.8"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/foobar167/articles/blob/master/Machine_Learning/code_examples/unet_segmentation.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "nDXW7A7pwZ7a",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# UNET SEGMENTATION"
      ]
    },
    {
      "metadata": {
        "id": "ofaEJYI4xBtD",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## [Original article](https://github.com/nikhilroxtomar/UNet-Segmentation-in-Keras-TensorFlow/blob/master/unet-segmentation.ipynb) of [Nikhil Tomar](https://github.com/nikhilroxtomar) on GitHub."
      ]
    },
    {
      "metadata": {
        "id": "LKmnuqTT_qbb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 421
        },
        "cellView": "form",
        "outputId": "fb0c2b9d-f21d-45cc-e59d-f24578fd58b3"
      },
      "cell_type": "code",
      "source": [
        "#@title Unet segmentation in Keras TensorFlow\n",
        "#@markdown Video explanation:\n",
        "from IPython.display import YouTubeVideo\n",
        "YouTubeVideo('M3EZS__Z_XE', width=600, height=400)\n"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "\n",
              "        <iframe\n",
              "            width=\"600\"\n",
              "            height=\"400\"\n",
              "            src=\"https://www.youtube.com/embed/M3EZS__Z_XE\"\n",
              "            frameborder=\"0\"\n",
              "            allowfullscreen\n",
              "        ></iframe>\n",
              "        "
            ],
            "text/plain": [
              "<IPython.lib.display.YouTubeVideo at 0x7f7f64fe2eb8>"
            ],
            "image/jpeg": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wCEAAUDBAgICAgICAgICAgICAgICAgICAgICAgICAgICAgI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\n"
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 1
        }
      ]
    },
    {
      "metadata": {
        "id": "KPJGmZNowZ7g",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Paper Arxiv Link: <a href=\"https://arxiv.org/abs/1505.04597\">U-Net: Convolutional Networks for Biomedical Image Segmentation</a>"
      ]
    },
    {
      "metadata": {
        "id": "ilY956O_wZ7j",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "<ul>\n",
        "<li>UNet is a fully convolutional network(FCN) that does image segmentation. Its goal is to predict each pixel's class.</li>\n",
        " \n",
        "<li>UNet is built upon the FCN and modified in a way that it yields better segmentation in medical imaging.</li>\n",
        "</ul>"
      ]
    },
    {
      "metadata": {
        "id": "mFWY7mLrwZ7l",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 1.1 Architecture\n",
        "\n",
        "<img src=\"https://raw.githubusercontent.com/foobar167/articles/master/Machine_Learning/data/2019.04.26_u-net-architecture.png\"/>\n",
        "\n",
        "<h3>UNet Architecture has 3 parts:</h3>\n",
        "<ol>\n",
        "    <li>The Contracting/Downsampling Path</li>\n",
        "    <li>Bottleneck</li>\n",
        "    <li>The Expanding/Upsampling Path</li>\n",
        "</ol>\n",
        "\n",
        "<h3>Downsampling Path: </h3> \n",
        "<ol>\n",
        "    <li>It consists of two 3x3 convolutions (unpadded convolutions), each followed by a rectified linear unit (ReLU) and a 2x2 max pooling operation with stride 2 for downsampling.</li> \n",
        "    <li>At each downsampling step we double the number of feature channels.</li>\n",
        "</ol>\n",
        "\n",
        "<h3>Upsampling Path: </h3> \n",
        "<ol>\n",
        "     <li> Every  step  in  the  expansive  path  consists  of  an  upsampling  of  the feature map followed by a 2x2 convolution (“up-convolution”), a concatenation with the correspondingly feature  map  from  the  downsampling  path,  and  two  3x3  convolutions,  each  followed by a ReLU.</li>\n",
        "</ol>\n",
        "\n",
        "<h3> Skip Connection: </h3>\n",
        "The skip connection from the downsampling path are concatenated with feature map during upsampling path. These skip connection provide local information to global information while upsampling.\n",
        "\n",
        "<h3> Final Layer: </h3>\n",
        "At the final layer a 1x1 convolution is used to map each feature vector to the desired number of classes."
      ]
    },
    {
      "metadata": {
        "id": "DkyvedSZwZ7s",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 1.2 Advantages\n",
        "<h3> Advantages: </h3>\n",
        "<ol>\n",
        "    <li>The UNet combines the location information from the downsampling path to finally obtain a general information combining localisation and context, which is necessary to predict a good segmentation map.</li>\n",
        "    <li>No Dense layer is used, so image sizes can be used.</li>\n",
        "</ol>"
      ]
    },
    {
      "metadata": {
        "id": "KK5ZNhS2wZ7t",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 1.3 Dataset\n",
        "Link: <a href=\"https://www.kaggle.com/c/data-science-bowl-2018\">Data Science Bowl 2018</a>\n",
        "Find the nuclei in divergent images to advance medical discovery"
      ]
    },
    {
      "metadata": {
        "id": "CSn4ToAHzRhB",
        "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": "OK"
            }
          },
          "base_uri": "https://localhost:8080/",
          "height": 77
        },
        "outputId": "9d5849c8-cd12-456a-f865-7de8a61e2678"
      },
      "cell_type": "code",
      "source": [
        "# Get file data-science-bowl-2018/stage1_train.zip\n",
        "# Download it from here: https://www.kaggle.com/c/8089/download-all\n",
        "# And upload stage1_train.zip to the Google Colab\n",
        "from google.colab import files\n",
        "uploaded = files.upload()"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-038220e9-1a5c-4db8-b67f-937b9d560f65\" name=\"files[]\" multiple disabled />\n",
              "     <output id=\"result-038220e9-1a5c-4db8-b67f-937b9d560f65\">\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 stage1_train.zip to stage1_train.zip\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "75VqBc8RDM4p",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Unzip it \n",
        "!mkdir stage1_train  # create directory\n",
        "!unzip stage1_train.zip -d stage1_train  # unzip into 'stage1_train' dir\n",
        "!rm stage1_train.zip  # delete 'stage1_train.zip' file"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "4tOIa8UJwZ7y",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## 1.4 Code"
      ]
    },
    {
      "metadata": {
        "id": "M5oF6qqZwZ70",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "## Imports\n",
        "import os\n",
        "import sys\n",
        "import random\n",
        "\n",
        "import numpy as np\n",
        "import cv2\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "## Seeding \n",
        "seed = 42\n",
        "random.seed = seed\n",
        "np.random.seed = seed\n",
        "tf.seed = seed"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "70jPovBLwZ7-",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Data Generator"
      ]
    },
    {
      "metadata": {
        "id": "SvRbQp8LwZ8B",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class DataGen(keras.utils.Sequence):\n",
        "    def __init__(self, ids, path, batch_size=8, image_size=128):\n",
        "        self.ids = ids\n",
        "        self.path = path\n",
        "        self.batch_size = batch_size\n",
        "        self.image_size = image_size\n",
        "        self.on_epoch_end()\n",
        "        \n",
        "    def __load__(self, id_name):\n",
        "        ## Path\n",
        "        image_path = os.path.join(self.path, id_name, \"images\", id_name) + \".png\"\n",
        "        mask_path = os.path.join(self.path, id_name, \"masks/\")\n",
        "        all_masks = os.listdir(mask_path)\n",
        "        \n",
        "        ## Reading Image\n",
        "        image = cv2.imread(image_path, 1)\n",
        "        image = cv2.resize(image, (self.image_size, self.image_size))\n",
        "        \n",
        "        mask = np.zeros((self.image_size, self.image_size, 1))\n",
        "        \n",
        "        ## Reading Masks\n",
        "        for name in all_masks:\n",
        "            _mask_path = mask_path + name\n",
        "            _mask_image = cv2.imread(_mask_path, -1)\n",
        "            _mask_image = cv2.resize(_mask_image, (self.image_size, self.image_size)) #128x128\n",
        "            _mask_image = np.expand_dims(_mask_image, axis=-1)\n",
        "            mask = np.maximum(mask, _mask_image)\n",
        "            \n",
        "        ## Normalizing \n",
        "        image = image/255.0\n",
        "        mask = mask/255.0\n",
        "        \n",
        "        return image, mask\n",
        "    \n",
        "    def __getitem__(self, index):\n",
        "        if(index+1)*self.batch_size > len(self.ids):\n",
        "            self.batch_size = len(self.ids) - index*self.batch_size\n",
        "        \n",
        "        files_batch = self.ids[index*self.batch_size : (index+1)*self.batch_size]\n",
        "        \n",
        "        image = []\n",
        "        mask  = []\n",
        "        \n",
        "        for id_name in files_batch:\n",
        "            _img, _mask = self.__load__(id_name)\n",
        "            image.append(_img)\n",
        "            mask.append(_mask)\n",
        "            \n",
        "        image = np.array(image)\n",
        "        mask  = np.array(mask)\n",
        "        \n",
        "        return image, mask\n",
        "    \n",
        "    def on_epoch_end(self):\n",
        "        pass\n",
        "    \n",
        "    def __len__(self):\n",
        "        return int(np.ceil(len(self.ids)/float(self.batch_size)))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "rUG6cDK1wZ8I",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Hyperparameters"
      ]
    },
    {
      "metadata": {
        "id": "2QwbWknPwZ8M",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "image_size = 128\n",
        "train_path = \"stage1_train/\"\n",
        "epochs = 10\n",
        "batch_size = 8\n",
        "\n",
        "## Training Ids\n",
        "train_ids = next(os.walk(train_path))[1]\n",
        "\n",
        "## Validation Data Size\n",
        "val_data_size = 10\n",
        "\n",
        "valid_ids = train_ids[:val_data_size]\n",
        "train_ids = train_ids[val_data_size:]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6_cMh29vwZ8R",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "0619de83-b019-4003-a63a-04c7289c3411"
      },
      "cell_type": "code",
      "source": [
        "gen = DataGen(train_ids, train_path, batch_size=batch_size, image_size=image_size)\n",
        "x, y = gen.__getitem__(0)\n",
        "print(x.shape, y.shape)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(8, 128, 128, 3) (8, 128, 128, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "scrolled": false,
        "id": "NCG2jWo9wZ8c",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 209
        },
        "outputId": "e53735ff-a588-4f68-9b6f-a2119c801d7a"
      },
      "cell_type": "code",
      "source": [
        "r = random.randint(0, len(x)-1)\n",
        "\n",
        "fig = plt.figure()\n",
        "fig.subplots_adjust(hspace=0.4, wspace=0.4)\n",
        "ax = fig.add_subplot(1, 2, 1)\n",
        "ax.imshow(x[r])\n",
        "ax = fig.add_subplot(1, 2, 2)\n",
        "ax.imshow(np.reshape(y[r], (image_size, image_size)), cmap=\"gray\")"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7f12a48e80>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAACuCAYAAAA4eMYdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDw+PDYcneg8PD48Nhyd6Dw8Pjw3H/w8Heg4L4OuJQQAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "1xJMvRqswZ8o",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Different Convolutional Blocks"
      ]
    },
    {
      "metadata": {
        "id": "xtA9Rr75wZ8p",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def down_block(x, filters, kernel_size=(3, 3), padding=\"same\", strides=1):\n",
        "    c = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation=\"relu\")(x)\n",
        "    c = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation=\"relu\")(c)\n",
        "    p = keras.layers.MaxPool2D((2, 2), (2, 2))(c)\n",
        "    return c, p\n",
        "\n",
        "def up_block(x, skip, filters, kernel_size=(3, 3), padding=\"same\", strides=1):\n",
        "    us = keras.layers.UpSampling2D((2, 2))(x)\n",
        "    concat = keras.layers.Concatenate()([us, skip])\n",
        "    c = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation=\"relu\")(concat)\n",
        "    c = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation=\"relu\")(c)\n",
        "    return c\n",
        "\n",
        "def bottleneck(x, filters, kernel_size=(3, 3), padding=\"same\", strides=1):\n",
        "    c = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation=\"relu\")(x)\n",
        "    c = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation=\"relu\")(c)\n",
        "    return c"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "QMdNQ_HVwZ82",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## UNet Model"
      ]
    },
    {
      "metadata": {
        "id": "YAa1RbB8wZ89",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def UNet():\n",
        "    f = [16, 32, 64, 128, 256]\n",
        "    inputs = keras.layers.Input((image_size, image_size, 3))\n",
        "    \n",
        "    p0 = inputs\n",
        "    c1, p1 = down_block(p0, f[0])  # 128 --> 64\n",
        "    c2, p2 = down_block(p1, f[1])  # 64  --> 32\n",
        "    c3, p3 = down_block(p2, f[2])  # 32  --> 16\n",
        "    c4, p4 = down_block(p3, f[3])  # 16  --> 8\n",
        "    \n",
        "    bn = bottleneck(p4, f[4])\n",
        "    \n",
        "    u1 = up_block(bn, c4, f[3])  # 8  --> 16\n",
        "    u2 = up_block(u1, c3, f[2])  # 16 --> 32\n",
        "    u3 = up_block(u2, c2, f[1])  # 32 --> 64\n",
        "    u4 = up_block(u3, c1, f[0])  # 64 --> 128\n",
        "    \n",
        "    outputs = keras.layers.Conv2D(1, (1, 1), padding=\"same\", activation=\"sigmoid\")(u4)\n",
        "    model = keras.models.Model(inputs, outputs)\n",
        "    return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "scrolled": true,
        "id": "ntBPY06awZ9E",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1470
        },
        "outputId": "e415feea-6985-4ba9-af26-b8038218c3a7"
      },
      "cell_type": "code",
      "source": [
        "model = UNet()\n",
        "model.compile(optimizer=\"adam\", loss=\"binary_crossentropy\", metrics=[\"acc\"])\n",
        "model.summary()"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Colocations handled automatically by placer.\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            (None, 128, 128, 3)  0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d (Conv2D)                 (None, 128, 128, 16) 448         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 128, 128, 16) 2320        conv2d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D)    (None, 64, 64, 16)   0           conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 64, 64, 32)   4640        max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 64, 64, 32)   9248        conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2D)  (None, 32, 32, 32)   0           conv2d_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_4 (Conv2D)               (None, 32, 32, 64)   18496       max_pooling2d_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_5 (Conv2D)               (None, 32, 32, 64)   36928       conv2d_4[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2D)  (None, 16, 16, 64)   0           conv2d_5[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_6 (Conv2D)               (None, 16, 16, 128)  73856       max_pooling2d_2[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_7 (Conv2D)               (None, 16, 16, 128)  147584      conv2d_6[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2D)  (None, 8, 8, 128)    0           conv2d_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_8 (Conv2D)               (None, 8, 8, 256)    295168      max_pooling2d_3[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_9 (Conv2D)               (None, 8, 8, 256)    590080      conv2d_8[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "up_sampling2d (UpSampling2D)    (None, 16, 16, 256)  0           conv2d_9[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "concatenate (Concatenate)       (None, 16, 16, 384)  0           up_sampling2d[0][0]              \n",
            "                                                                 conv2d_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_10 (Conv2D)              (None, 16, 16, 128)  442496      concatenate[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_11 (Conv2D)              (None, 16, 16, 128)  147584      conv2d_10[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "up_sampling2d_1 (UpSampling2D)  (None, 32, 32, 128)  0           conv2d_11[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_1 (Concatenate)     (None, 32, 32, 192)  0           up_sampling2d_1[0][0]            \n",
            "                                                                 conv2d_5[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_12 (Conv2D)              (None, 32, 32, 64)   110656      concatenate_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_13 (Conv2D)              (None, 32, 32, 64)   36928       conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "up_sampling2d_2 (UpSampling2D)  (None, 64, 64, 64)   0           conv2d_13[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_2 (Concatenate)     (None, 64, 64, 96)   0           up_sampling2d_2[0][0]            \n",
            "                                                                 conv2d_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_14 (Conv2D)              (None, 64, 64, 32)   27680       concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_15 (Conv2D)              (None, 64, 64, 32)   9248        conv2d_14[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "up_sampling2d_3 (UpSampling2D)  (None, 128, 128, 32) 0           conv2d_15[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_3 (Concatenate)     (None, 128, 128, 48) 0           up_sampling2d_3[0][0]            \n",
            "                                                                 conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_16 (Conv2D)              (None, 128, 128, 16) 6928        concatenate_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_17 (Conv2D)              (None, 128, 128, 16) 2320        conv2d_16[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_18 (Conv2D)              (None, 128, 128, 1)  17          conv2d_17[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 1,962,625\n",
            "Trainable params: 1,962,625\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Nmkxq4-BwZ9Q",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Training the model"
      ]
    },
    {
      "metadata": {
        "scrolled": false,
        "id": "6I34e4tcwZ9T",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 661
        },
        "outputId": "58622df7-5cfb-402e-9caa-842caeb4ac83"
      },
      "cell_type": "code",
      "source": [
        "train_gen = DataGen(train_ids, train_path, image_size=image_size, batch_size=batch_size)\n",
        "valid_gen = DataGen(valid_ids, train_path, image_size=image_size, batch_size=batch_size)\n",
        "\n",
        "train_steps = len(train_ids)//batch_size\n",
        "valid_steps = len(valid_ids)//batch_size\n",
        "\n",
        "model.fit_generator(train_gen, validation_data=valid_gen, steps_per_epoch=train_steps,\n",
        "                    validation_steps=valid_steps, epochs=epochs)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n",
            "Epoch 1/10\n",
            "2/2 [==============================] - 1s 462ms/step - loss: 0.8926 - acc: 0.9346\n",
            "83/83 [==============================] - 24s 291ms/step - loss: 2.6685 - acc: 0.8320 - val_loss: 0.8926 - val_acc: 0.9346\n",
            "Epoch 2/10\n",
            "5/5 [==============================] - 0s 76ms/step - loss: 0.2497 - acc: 0.9107\n",
            "83/83 [==============================] - 14s 170ms/step - loss: 0.8620 - acc: 0.8629 - val_loss: 0.2497 - val_acc: 0.9107\n",
            "Epoch 3/10\n",
            "5/5 [==============================] - 0s 85ms/step - loss: 0.1496 - acc: 0.9668\n",
            "83/83 [==============================] - 15s 177ms/step - loss: 0.6539 - acc: 0.9156 - val_loss: 0.1496 - val_acc: 0.9668\n",
            "Epoch 4/10\n",
            "5/5 [==============================] - 0s 79ms/step - loss: 0.1137 - acc: 0.9709\n",
            "83/83 [==============================] - 15s 179ms/step - loss: 0.4919 - acc: 0.9382 - val_loss: 0.1137 - val_acc: 0.9709\n",
            "Epoch 5/10\n",
            "5/5 [==============================] - 0s 78ms/step - loss: 0.1407 - acc: 0.9667\n",
            "83/83 [==============================] - 14s 172ms/step - loss: 0.5270 - acc: 0.9361 - val_loss: 0.1407 - val_acc: 0.9667\n",
            "Epoch 6/10\n",
            "5/5 [==============================] - 0s 83ms/step - loss: 0.1272 - acc: 0.9680\n",
            "83/83 [==============================] - 14s 172ms/step - loss: 0.4950 - acc: 0.9378 - val_loss: 0.1272 - val_acc: 0.9680\n",
            "Epoch 7/10\n",
            "5/5 [==============================] - 0s 79ms/step - loss: 0.1142 - acc: 0.9684\n",
            "83/83 [==============================] - 14s 172ms/step - loss: 0.5161 - acc: 0.9344 - val_loss: 0.1142 - val_acc: 0.9684\n",
            "Epoch 8/10\n",
            "5/5 [==============================] - 0s 80ms/step - loss: 0.0952 - acc: 0.9717\n",
            "83/83 [==============================] - 15s 175ms/step - loss: 0.4100 - acc: 0.9449 - val_loss: 0.0952 - val_acc: 0.9717\n",
            "Epoch 9/10\n",
            "5/5 [==============================] - 0s 73ms/step - loss: 0.0981 - acc: 0.9712\n",
            "83/83 [==============================] - 16s 188ms/step - loss: 0.3795 - acc: 0.9478 - val_loss: 0.0981 - val_acc: 0.9712\n",
            "Epoch 10/10\n",
            "5/5 [==============================] - 0s 82ms/step - loss: 0.0883 - acc: 0.9726\n",
            "83/83 [==============================] - 14s 172ms/step - loss: 0.3781 - acc: 0.9481 - val_loss: 0.0883 - val_acc: 0.9726\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f7f0e253ba8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "metadata": {
        "id": "iDUykT_awZ9h",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Testing the model"
      ]
    },
    {
      "metadata": {
        "id": "i_oY4H3LwZ9l",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "## Save the Weights\n",
        "model.save_weights(\"UNetW.h5\")\n",
        "\n",
        "## Dataset for prediction\n",
        "x, y = valid_gen.__getitem__(2)\n",
        "result = model.predict(x)\n",
        "\n",
        "result = result > 0.5"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "uSdH42qVwZ9q",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 209
        },
        "outputId": "e20e9523-d4b2-441c-8e36-4a1a0fdde8b9"
      },
      "cell_type": "code",
      "source": [
        "fig = plt.figure()\n",
        "fig.subplots_adjust(hspace=0.4, wspace=0.4)\n",
        "\n",
        "ax = fig.add_subplot(1, 2, 1)\n",
        "ax.imshow(np.reshape(y[0]*255, (image_size, image_size)), cmap=\"gray\")\n",
        "\n",
        "ax = fig.add_subplot(1, 2, 2)\n",
        "ax.imshow(np.reshape(result[0]*255, (image_size, image_size)), cmap=\"gray\")"
      ],
      "execution_count": 56,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7f0ac86940>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 56
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAACuCAYAAAA4eMYdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\naiAAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "1OKwJGbVwZ-A",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 209
        },
        "outputId": "60c678b7-9fa9-49d3-dac9-17c95b4cf9eb"
      },
      "cell_type": "code",
      "source": [
        "fig = plt.figure()\n",
        "fig.subplots_adjust(hspace=0.4, wspace=0.4)\n",
        "\n",
        "ax = fig.add_subplot(1, 2, 1)\n",
        "ax.imshow(np.reshape(y[1]*255, (image_size, image_size)), cmap=\"gray\")\n",
        "\n",
        "ax = fig.add_subplot(1, 2, 2)\n",
        "ax.imshow(np.reshape(result[1]*255, (image_size, image_size)), cmap=\"gray\")"
      ],
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f7f0ac273c8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 57
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAACuCAYAAAA4eMYdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "3gx211wTvgn2",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}