BUS_Tumor_Image_Segmentation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "BUS_Tumor_Image_Segmentation.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "# An Experimental Study on Breast Ultrasound Tumor Segementation\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "SCzQ-3ncO5bf"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "In this Notebook we explore the UNet based segmentation of Breast US Tumor Images. The main idea is to study the efficiency of automatic semantic segmentation of the images with the help of UNet."
      ],
      "metadata": {
        "id": "Nkre7UGbPJhF"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "For the purpose experimentation two publicly available datasets are utilized, \n",
        "\n",
        "*   BUSI dataset. (https://www.sciencedirect.com/science/article/pii/S2352340919312181)\n",
        "*   UDIAT dataset. (https://ieeexplore.ieee.org/abstract/document/8003418)\n",
        "\n"
      ],
      "metadata": {
        "id": "VkZPzDQ5PiVM"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "The dataset images have been preprocessed and converted to 224x224 dimensionality and saved as .npy files. These files will be used directly in this Notebook."
      ],
      "metadata": {
        "id": "iDLq6rqWQ7EA"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yCVfSODDNmHx",
        "outputId": "1c94a100-542d-4716-d809-17c52ece846e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/gdrive\n"
          ]
        }
      ],
      "source": [
        "# Connect to Google Drive to access dataset\n",
        "from google.colab import drive\n",
        "\n",
        "drive.mount('/content/gdrive')"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Check dataset filenames\n",
        "import os\n",
        "import cv2\n",
        "\n",
        "bus_filenames = os.listdir('/content/gdrive/My Drive/BreastUSDatasetNPYs')\n",
        "print(bus_filenames)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4SldMnNsOj0b",
        "outputId": "4efe818a-1528-4952-e295-1e876a430a22"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['busiTestImages.npy', 'busiTestLabels.npy', 'busiTestMasks.npy', 'busiTrainImages.npy', 'udTrainImages.npy', 'busiTrainMasks.npy', 'udTestLabels.npy', 'udTestMasks.npy', 'udTestImages.npy', 'busiTrainLabels.npy', 'udTrainLabels.npy', 'udTrainMasks.npy']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Read the Numpy files containing the dataset.\n",
        "import numpy as np\n",
        "\n",
        "busiTrainImages = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/busiTrainImages.npy')\n",
        "busiTrainMasks = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/busiTrainMasks.npy')\n",
        "busiTrainLabels = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/busiTrainLabels.npy')\n",
        "busiTestImages = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/busiTestImages.npy')\n",
        "busiTestMasks = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/busiTestMasks.npy')\n",
        "busiTestLabels = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/busiTestLabels.npy')\n",
        "\n",
        "udTrainImages = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/udTrainImages.npy')\n",
        "udTrainMasks = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/udTrainMasks.npy')\n",
        "udTrainLabels = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/udTrainLabels.npy')\n",
        "udTestImages = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/udTestImages.npy')\n",
        "udTestMasks = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/udTestMasks.npy')\n",
        "udTestLabels = np.load('/content/gdrive/My Drive/BreastUSDatasetNPYs/udTestLabels.npy')\n",
        "\n",
        "print(busiTrainImages.shape)\n",
        "print(busiTrainMasks.shape)\n",
        "print(busiTrainLabels.shape)\n",
        "print(busiTestImages.shape)\n",
        "print(busiTestMasks.shape)\n",
        "print(busiTestLabels.shape)\n",
        "\n",
        "print(udTrainImages.shape)\n",
        "print(udTrainMasks.shape)\n",
        "print(udTrainLabels.shape)\n",
        "print(udTestImages.shape)\n",
        "print(udTestMasks.shape)\n",
        "print(udTestLabels.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yl9jPekIO0JK",
        "outputId": "42994c34-f559-4a81-c8d3-d83f867121c0"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(532, 224, 224)\n",
            "(532, 224, 224)\n",
            "(532,)\n",
            "(133, 224, 224)\n",
            "(133, 224, 224)\n",
            "(133,)\n",
            "(130, 224, 224)\n",
            "(130, 224, 224)\n",
            "(130,)\n",
            "(33, 224, 224)\n",
            "(33, 224, 224)\n",
            "(33,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Inspect the pixel value range \n",
        "\n",
        "print(busiTrainImages[0])\n",
        "print(busiTrainMasks[0])\n",
        "print(busiTrainLabels[0])\n",
        "\n",
        "# The images and also the masks are having pixel values in the range 0-255\n",
        "# We need to rescale the data to bring all pixel values to the range 0 to 1\n",
        "\n",
        "busiTrainImages = busiTrainImages/255.\n",
        "busiTrainMasks = busiTrainMasks/255.\n",
        "\n",
        "busiTestImages = busiTestImages/255.\n",
        "busiTestMasks = busiTestMasks/255.\n",
        "\n",
        "udTrainImages = udTrainImages/255.\n",
        "udTrainMasks = udTrainMasks/255.\n",
        "\n",
        "udTestImages = udTestImages/255.\n",
        "udTestMasks = udTestMasks/255."
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tictvS2_R8W8",
        "outputId": "3b6d440b-445d-448b-f79f-fefbf122febb"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[ 80 174  37 ... 208 207 199]\n",
            " [ 83 207  21 ... 195 199 202]\n",
            " [ 85 161  42 ... 180 182 186]\n",
            " ...\n",
            " [ 20  19  16 ...  36  46  54]\n",
            " [ 29  27  23 ...  57  53  46]\n",
            " [ 20  20  23 ...  37  32  30]]\n",
            "[[0 0 0 ... 0 0 0]\n",
            " [0 0 0 ... 0 0 0]\n",
            " [0 0 0 ... 0 0 0]\n",
            " ...\n",
            " [0 0 0 ... 0 0 0]\n",
            " [0 0 0 ... 0 0 0]\n",
            " [0 0 0 ... 0 0 0]]\n",
            "0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# After rescaling the pixel values are modified\n",
        "print(busiTrainImages[0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0i-hlywsTh_3",
        "outputId": "54632d81-fb60-4f53-f73c-d94357b8f54c"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[0.31372549 0.68235294 0.14509804 ... 0.81568627 0.81176471 0.78039216]\n",
            " [0.3254902  0.81176471 0.08235294 ... 0.76470588 0.78039216 0.79215686]\n",
            " [0.33333333 0.63137255 0.16470588 ... 0.70588235 0.71372549 0.72941176]\n",
            " ...\n",
            " [0.07843137 0.0745098  0.0627451  ... 0.14117647 0.18039216 0.21176471]\n",
            " [0.11372549 0.10588235 0.09019608 ... 0.22352941 0.20784314 0.18039216]\n",
            " [0.07843137 0.07843137 0.09019608 ... 0.14509804 0.1254902  0.11764706]]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(busiTrainImages.shape)\n",
        "print(busiTrainMasks.shape)\n",
        "print(udTrainImages.shape)\n",
        "print(udTrainMasks.shape)\n",
        "print(busiTestImages.shape)\n",
        "print(busiTestMasks.shape)\n",
        "print(udTestImages.shape)\n",
        "print(udTestMasks.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P-1JbGYXB4WQ",
        "outputId": "49c081a3-9f06-4056-85e4-107e0461f6ff"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(532, 224, 224)\n",
            "(532, 224, 224)\n",
            "(130, 224, 224)\n",
            "(130, 224, 224)\n",
            "(133, 224, 224)\n",
            "(133, 224, 224)\n",
            "(33, 224, 224)\n",
            "(33, 224, 224)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Reshape all the samples into (#samples,224,224,1)\n",
        "\n",
        "busiTrainImages = busiTrainImages.reshape((-1, 224, 224, 1))\n",
        "busiTrainMasks = busiTrainMasks.reshape((-1, 224, 224, 1))\n",
        "busiTestImages = busiTestImages.reshape((-1, 224, 224, 1))\n",
        "busiTestMasks = busiTestMasks.reshape((-1, 224, 224, 1))\n",
        "\n",
        "udTrainImages = udTrainImages.reshape((-1, 224, 224, 1))\n",
        "udTrainMasks = udTrainMasks.reshape((-1, 224, 224, 1))\n",
        "udTestImages = udTestImages.reshape((-1, 224, 224, 1))\n",
        "udTestMasks = udTestMasks.reshape((-1, 224, 224, 1))\n",
        "\n",
        "print(busiTrainImages.shape)\n",
        "print(busiTrainMasks.shape)\n",
        "print(udTrainImages.shape)\n",
        "print(udTrainMasks.shape)\n",
        "print(busiTestImages.shape)\n",
        "print(busiTestMasks.shape)\n",
        "print(udTestImages.shape)\n",
        "print(udTestMasks.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lmUR9A6MgTFo",
        "outputId": "7bc115dd-1831-4447-8d0d-84de0570ddaa"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(532, 224, 224, 1)\n",
            "(532, 224, 224, 1)\n",
            "(130, 224, 224, 1)\n",
            "(130, 224, 224, 1)\n",
            "(133, 224, 224, 1)\n",
            "(133, 224, 224, 1)\n",
            "(33, 224, 224, 1)\n",
            "(33, 224, 224, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Since for the purpose of segmentation the number of samples is quite less, we combine the two datasets.\n",
        "combinedTrainImages = np.concatenate((busiTrainImages,udTrainImages),axis=0)\n",
        "combinedTrainMasks = np.concatenate((busiTrainMasks,udTrainMasks),axis=0)\n",
        "combinedTestImages = np.concatenate((busiTestImages,udTestImages),axis=0)\n",
        "combinedTestMasks = np.concatenate((busiTestMasks,udTestMasks),axis=0)\n",
        "\n",
        "print(combinedTrainImages.shape)\n",
        "print(combinedTrainMasks.shape)\n",
        "print(combinedTestImages.shape)\n",
        "print(combinedTestMasks.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IisjjclbboQU",
        "outputId": "8d80fe37-d9fa-43eb-8d6a-9c98ecfa44e0"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(662, 224, 224, 1)\n",
            "(662, 224, 224, 1)\n",
            "(166, 224, 224, 1)\n",
            "(166, 224, 224, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Split the training data into training and validation sets.\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "\n",
        "# Train test split\n",
        "\n",
        "xTrain,xVal,yTrain,yVal = train_test_split(combinedTrainImages,combinedTrainMasks, test_size=0.20, shuffle = True)\n",
        "\n",
        "\n",
        "print(xTrain.shape)\n",
        "print(yTrain.shape)\n",
        "print(xVal.shape)\n",
        "print(yVal.shape)\n",
        "print(combinedTestImages.shape)\n",
        "print(combinedTestMasks.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rY6JWiasdTZw",
        "outputId": "03fd682c-0ee4-45e0-e2e6-4c0976e84b4b"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(529, 224, 224, 1)\n",
            "(529, 224, 224, 1)\n",
            "(133, 224, 224, 1)\n",
            "(133, 224, 224, 1)\n",
            "(166, 224, 224, 1)\n",
            "(166, 224, 224, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Once our data is ready, next we prepare our model and loss functions.\n"
      ],
      "metadata": {
        "id": "EJ85BdWneXFQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import warnings\n",
        "from tensorflow.keras import backend as K\n",
        "from PIL import Image\n",
        "from tensorflow.keras.callbacks import EarlyStopping\n",
        "from tensorflow.keras.layers import Dense, Activation, Dropout, Flatten,Conv2D, MaxPooling2D\n",
        "from tensorflow.keras.layers import BatchNormalization\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from tensorflow.keras.layers import Input\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.layers import GlobalAveragePooling2D as GAP\n",
        "\n",
        "from keras.losses import binary_crossentropy\n",
        "import keras.backend as K\n",
        "\n",
        "\n",
        "def dice_coef(y_true, y_pred):\n",
        "    y_true_f = K.flatten(y_true)\n",
        "    y_pred = K.cast(y_pred, 'float32')\n",
        "    y_pred_f = K.cast(K.greater(K.flatten(y_pred), 0.5), 'float32')\n",
        "    intersection = y_true_f * y_pred_f\n",
        "    score = 2. * K.sum(intersection) / (K.sum(y_true_f) + K.sum(y_pred_f))\n",
        "    return score\n",
        "\n",
        "def dice_loss(y_true, y_pred):\n",
        "    y_true = K.cast(y_true, 'float32')\n",
        "    y_pred = K.cast(y_pred, 'float32')\n",
        "    smooth = 1.\n",
        "    y_true_f = K.flatten(y_true)\n",
        "    y_pred_f = K.flatten(y_pred)\n",
        "    intersection = y_true_f * y_pred_f\n",
        "    score = (2. * K.sum(intersection) + smooth) / (K.sum(y_true_f * y_true_f) + K.sum(y_pred_f * y_pred_f) + smooth)\n",
        "    return 1. - score\n",
        "\n",
        "def bce_dice_loss(y_true, y_pred):\n",
        "    y_true = K.cast(y_true, 'float32')\n",
        "    y_pred = K.cast(y_pred, 'float32')\n",
        "    return binary_crossentropy(y_true, y_pred) + dice_loss(y_true, y_pred)\n"
      ],
      "metadata": {
        "id": "Q7al6MjweVOw"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras.layers import Conv2DTranspose, UpSampling2D\n",
        "from tensorflow.keras.layers import *\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras.layers import GlobalAveragePooling2D as GAP\n",
        "from tensorflow.keras import layers\n",
        "\n",
        "IMG_HEIGHT=224\n",
        "IMG_WIDTH = 224\n",
        "IMG_CHANNELS = 1\n",
        "inputs = Input((IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS))\n",
        "\n",
        "c1 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (inputs)\n",
        "c1 = Dropout(0.1) (c1)\n",
        "c1 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c1)\n",
        "p1 = MaxPooling2D((2, 2)) (c1)\n",
        "c2 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p1)\n",
        "c2 = Dropout(0.1) (c2)\n",
        "c2 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c2)\n",
        "p2 = MaxPooling2D((2, 2)) (c2)\n",
        "c3 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p2)\n",
        "c3 = Dropout(0.2) (c3)\n",
        "c3 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c3)\n",
        "p3 = MaxPooling2D((2, 2)) (c3)\n",
        "c4 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p3)\n",
        "c4 = Dropout(0.2) (c4)\n",
        "c4 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c4)\n",
        "p4 = MaxPooling2D(pool_size=(2, 2)) (c4)\n",
        "c5 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p4)\n",
        "c5 = Dropout(0.3) (c5)\n",
        "c5 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c5)\n",
        "\n",
        "# Break here for multitask learning\n",
        "\n",
        "u6 = Conv2DTranspose(512,(2, 2), strides=(2, 2),padding='same') (c5)\n",
        "u6 = layers.concatenate([u6, c4])\n",
        "c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u6)\n",
        "c6 = Dropout(0.2) (c6)\n",
        "c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c6)\n",
        "u7 = Conv2DTranspose(256,(2, 2), strides=(2, 2),padding='same') (c6)\n",
        "u7 = layers.concatenate([u7, c3])\n",
        "c7 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u7)\n",
        "c7 = Dropout(0.2) (c7)\n",
        "c7 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c7)\n",
        "u8 = Conv2DTranspose(128,(2, 2), strides=(2, 2),padding='same') (c7)\n",
        "u8 = layers.concatenate([u8, c2])\n",
        "c8 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u8)\n",
        "c8 = Dropout(0.1) (c8)\n",
        "c8 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c8)\n",
        "u9 = Conv2DTranspose(64,(2, 2), strides=(2, 2),padding='same') (c8)\n",
        "u9 = layers.concatenate([u9, c1], axis=3)\n",
        "c9 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u9)\n",
        "c9 = Dropout(0.1) (c9)\n",
        "c9 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c9)\n",
        "seg_branch = Conv2D(1, (1, 1), activation='sigmoid', name='seg_output') (c9)\n",
        "\n",
        "model = tf.keras.Model(inputs = inputs, outputs = seg_branch)\n",
        "print(model.summary())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iH2ReHegffSk",
        "outputId": "df709fe0-9c57-4ccd-cf41-098aef02f87e"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)           [(None, 224, 224, 1  0           []                               \n",
            "                                )]                                                                \n",
            "                                                                                                  \n",
            " conv2d (Conv2D)                (None, 224, 224, 64  640         ['input_1[0][0]']                \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " dropout (Dropout)              (None, 224, 224, 64  0           ['conv2d[0][0]']                 \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_1 (Conv2D)              (None, 224, 224, 64  36928       ['dropout[0][0]']                \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " max_pooling2d (MaxPooling2D)   (None, 112, 112, 64  0           ['conv2d_1[0][0]']               \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_2 (Conv2D)              (None, 112, 112, 12  73856       ['max_pooling2d[0][0]']          \n",
            "                                8)                                                                \n",
            "                                                                                                  \n",
            " dropout_1 (Dropout)            (None, 112, 112, 12  0           ['conv2d_2[0][0]']               \n",
            "                                8)                                                                \n",
            "                                                                                                  \n",
            " conv2d_3 (Conv2D)              (None, 112, 112, 12  147584      ['dropout_1[0][0]']              \n",
            "                                8)                                                                \n",
            "                                                                                                  \n",
            " max_pooling2d_1 (MaxPooling2D)  (None, 56, 56, 128)  0          ['conv2d_3[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_4 (Conv2D)              (None, 56, 56, 256)  295168      ['max_pooling2d_1[0][0]']        \n",
            "                                                                                                  \n",
            " dropout_2 (Dropout)            (None, 56, 56, 256)  0           ['conv2d_4[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_5 (Conv2D)              (None, 56, 56, 256)  590080      ['dropout_2[0][0]']              \n",
            "                                                                                                  \n",
            " max_pooling2d_2 (MaxPooling2D)  (None, 28, 28, 256)  0          ['conv2d_5[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_6 (Conv2D)              (None, 28, 28, 512)  1180160     ['max_pooling2d_2[0][0]']        \n",
            "                                                                                                  \n",
            " dropout_3 (Dropout)            (None, 28, 28, 512)  0           ['conv2d_6[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_7 (Conv2D)              (None, 28, 28, 512)  2359808     ['dropout_3[0][0]']              \n",
            "                                                                                                  \n",
            " max_pooling2d_3 (MaxPooling2D)  (None, 14, 14, 512)  0          ['conv2d_7[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_8 (Conv2D)              (None, 14, 14, 1024  4719616     ['max_pooling2d_3[0][0]']        \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " dropout_4 (Dropout)            (None, 14, 14, 1024  0           ['conv2d_8[0][0]']               \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_9 (Conv2D)              (None, 14, 14, 1024  9438208     ['dropout_4[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_transpose (Conv2DTransp  (None, 28, 28, 512)  2097664    ['conv2d_9[0][0]']               \n",
            " ose)                                                                                             \n",
            "                                                                                                  \n",
            " concatenate (Concatenate)      (None, 28, 28, 1024  0           ['conv2d_transpose[0][0]',       \n",
            "                                )                                 'conv2d_7[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_10 (Conv2D)             (None, 28, 28, 512)  4719104     ['concatenate[0][0]']            \n",
            "                                                                                                  \n",
            " dropout_5 (Dropout)            (None, 28, 28, 512)  0           ['conv2d_10[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_11 (Conv2D)             (None, 28, 28, 512)  2359808     ['dropout_5[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_transpose_1 (Conv2DTran  (None, 56, 56, 256)  524544     ['conv2d_11[0][0]']              \n",
            " spose)                                                                                           \n",
            "                                                                                                  \n",
            " concatenate_1 (Concatenate)    (None, 56, 56, 512)  0           ['conv2d_transpose_1[0][0]',     \n",
            "                                                                  'conv2d_5[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_12 (Conv2D)             (None, 56, 56, 256)  1179904     ['concatenate_1[0][0]']          \n",
            "                                                                                                  \n",
            " dropout_6 (Dropout)            (None, 56, 56, 256)  0           ['conv2d_12[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_13 (Conv2D)             (None, 56, 56, 256)  590080      ['dropout_6[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_transpose_2 (Conv2DTran  (None, 112, 112, 12  131200     ['conv2d_13[0][0]']              \n",
            " spose)                         8)                                                                \n",
            "                                                                                                  \n",
            " concatenate_2 (Concatenate)    (None, 112, 112, 25  0           ['conv2d_transpose_2[0][0]',     \n",
            "                                6)                                'conv2d_3[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_14 (Conv2D)             (None, 112, 112, 12  295040      ['concatenate_2[0][0]']          \n",
            "                                8)                                                                \n",
            "                                                                                                  \n",
            " dropout_7 (Dropout)            (None, 112, 112, 12  0           ['conv2d_14[0][0]']              \n",
            "                                8)                                                                \n",
            "                                                                                                  \n",
            " conv2d_15 (Conv2D)             (None, 112, 112, 12  147584      ['dropout_7[0][0]']              \n",
            "                                8)                                                                \n",
            "                                                                                                  \n",
            " conv2d_transpose_3 (Conv2DTran  (None, 224, 224, 64  32832      ['conv2d_15[0][0]']              \n",
            " spose)                         )                                                                 \n",
            "                                                                                                  \n",
            " concatenate_3 (Concatenate)    (None, 224, 224, 12  0           ['conv2d_transpose_3[0][0]',     \n",
            "                                8)                                'conv2d_1[0][0]']               \n",
            "                                                                                                  \n",
            " conv2d_16 (Conv2D)             (None, 224, 224, 64  73792       ['concatenate_3[0][0]']          \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " dropout_8 (Dropout)            (None, 224, 224, 64  0           ['conv2d_16[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_17 (Conv2D)             (None, 224, 224, 64  36928       ['dropout_8[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " seg_output (Conv2D)            (None, 224, 224, 1)  65          ['conv2d_17[0][0]']              \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 31,030,593\n",
            "Trainable params: 31,030,593\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n",
            "None\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Compile model\n",
        "\n",
        "model.compile(optimizer='adam', loss=bce_dice_loss, metrics=[dice_coef])"
      ],
      "metadata": {
        "id": "wJVOJ6PGlgCL"
      },
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Model training with earlystopping\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from tensorflow.keras.callbacks import EarlyStopping\n",
        "\n",
        "es = EarlyStopping(monitor='val_loss', patience=100, verbose=1, mode='min',restore_best_weights=True)\n",
        "\n",
        "history = model.fit(xTrain, yTrain,\n",
        "              epochs=500,\n",
        "              verbose=1,\n",
        "              validation_data = (xVal, yVal), batch_size=16,\n",
        "              callbacks = [es]\n",
        "             )\n",
        "\n",
        "# Plot the training and validation losses\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "oreG-Jv2l5Ih",
        "outputId": "19a8ee74-138f-492a-845c-21917b4d77a3"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/500\n",
            "34/34 [==============================] - 71s 1s/step - loss: 1.4232 - dice_coef: 0.1051 - val_loss: 1.2062 - val_dice_coef: 0.0039\n",
            "Epoch 2/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 1.0374 - dice_coef: 0.1463 - val_loss: 1.1272 - val_dice_coef: 0.0457\n",
            "Epoch 3/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.9529 - dice_coef: 0.2104 - val_loss: 0.8753 - val_dice_coef: 0.2548\n",
            "Epoch 4/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.9018 - dice_coef: 0.2921 - val_loss: 0.8561 - val_dice_coef: 0.3697\n",
            "Epoch 5/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.8359 - dice_coef: 0.3175 - val_loss: 0.7823 - val_dice_coef: 0.3752\n",
            "Epoch 6/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.8036 - dice_coef: 0.3408 - val_loss: 0.8125 - val_dice_coef: 0.3165\n",
            "Epoch 7/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.8097 - dice_coef: 0.3523 - val_loss: 0.7869 - val_dice_coef: 0.4109\n",
            "Epoch 8/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.7751 - dice_coef: 0.3962 - val_loss: 0.7502 - val_dice_coef: 0.4233\n",
            "Epoch 9/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.7390 - dice_coef: 0.4096 - val_loss: 0.7917 - val_dice_coef: 0.2394\n",
            "Epoch 10/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.7129 - dice_coef: 0.4391 - val_loss: 0.7947 - val_dice_coef: 0.2829\n",
            "Epoch 11/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6872 - dice_coef: 0.4497 - val_loss: 0.6953 - val_dice_coef: 0.4420\n",
            "Epoch 12/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6937 - dice_coef: 0.4552 - val_loss: 0.8030 - val_dice_coef: 0.2249\n",
            "Epoch 13/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6595 - dice_coef: 0.4666 - val_loss: 0.6648 - val_dice_coef: 0.4609\n",
            "Epoch 14/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6606 - dice_coef: 0.4953 - val_loss: 0.7630 - val_dice_coef: 0.3711\n",
            "Epoch 15/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6428 - dice_coef: 0.5056 - val_loss: 0.9545 - val_dice_coef: 0.2376\n",
            "Epoch 16/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.7077 - dice_coef: 0.4524 - val_loss: 0.6957 - val_dice_coef: 0.4553\n",
            "Epoch 17/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6636 - dice_coef: 0.4759 - val_loss: 0.7279 - val_dice_coef: 0.4125\n",
            "Epoch 18/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6376 - dice_coef: 0.5166 - val_loss: 0.6134 - val_dice_coef: 0.5354\n",
            "Epoch 19/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.6269 - dice_coef: 0.5233 - val_loss: 0.6677 - val_dice_coef: 0.4638\n",
            "Epoch 20/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.6505 - dice_coef: 0.4912 - val_loss: 0.6380 - val_dice_coef: 0.4799\n",
            "Epoch 21/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.6175 - dice_coef: 0.5145 - val_loss: 0.7055 - val_dice_coef: 0.4092\n",
            "Epoch 22/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.6247 - dice_coef: 0.5060 - val_loss: 0.8909 - val_dice_coef: 0.2970\n",
            "Epoch 23/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5958 - dice_coef: 0.5178 - val_loss: 0.6064 - val_dice_coef: 0.5264\n",
            "Epoch 24/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5685 - dice_coef: 0.5529 - val_loss: 0.6944 - val_dice_coef: 0.4058\n",
            "Epoch 25/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5966 - dice_coef: 0.5330 - val_loss: 0.5876 - val_dice_coef: 0.5404\n",
            "Epoch 26/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5565 - dice_coef: 0.5537 - val_loss: 0.6319 - val_dice_coef: 0.5354\n",
            "Epoch 27/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5850 - dice_coef: 0.5666 - val_loss: 0.6808 - val_dice_coef: 0.4686\n",
            "Epoch 28/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5606 - dice_coef: 0.5549 - val_loss: 0.5733 - val_dice_coef: 0.5602\n",
            "Epoch 29/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.5313 - dice_coef: 0.5731 - val_loss: 0.5936 - val_dice_coef: 0.5620\n",
            "Epoch 30/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5203 - dice_coef: 0.5913 - val_loss: 0.5657 - val_dice_coef: 0.5502\n",
            "Epoch 31/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.5358 - dice_coef: 0.6021 - val_loss: 0.6228 - val_dice_coef: 0.4845\n",
            "Epoch 32/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.5382 - dice_coef: 0.5932 - val_loss: 0.5344 - val_dice_coef: 0.6036\n",
            "Epoch 33/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4835 - dice_coef: 0.6387 - val_loss: 0.5722 - val_dice_coef: 0.5587\n",
            "Epoch 34/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4718 - dice_coef: 0.6260 - val_loss: 0.5573 - val_dice_coef: 0.5632\n",
            "Epoch 35/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4833 - dice_coef: 0.6371 - val_loss: 0.5800 - val_dice_coef: 0.5609\n",
            "Epoch 36/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4483 - dice_coef: 0.6641 - val_loss: 0.5828 - val_dice_coef: 0.5652\n",
            "Epoch 37/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4598 - dice_coef: 0.6596 - val_loss: 0.5579 - val_dice_coef: 0.5687\n",
            "Epoch 38/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4475 - dice_coef: 0.6673 - val_loss: 0.5785 - val_dice_coef: 0.5407\n",
            "Epoch 39/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4750 - dice_coef: 0.6432 - val_loss: 0.5629 - val_dice_coef: 0.5948\n",
            "Epoch 40/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4385 - dice_coef: 0.6730 - val_loss: 0.5262 - val_dice_coef: 0.6184\n",
            "Epoch 41/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4121 - dice_coef: 0.6959 - val_loss: 0.5056 - val_dice_coef: 0.6327\n",
            "Epoch 42/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4072 - dice_coef: 0.6956 - val_loss: 0.5033 - val_dice_coef: 0.6384\n",
            "Epoch 43/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3973 - dice_coef: 0.7090 - val_loss: 0.5491 - val_dice_coef: 0.6090\n",
            "Epoch 44/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4134 - dice_coef: 0.6923 - val_loss: 0.5139 - val_dice_coef: 0.6290\n",
            "Epoch 45/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4262 - dice_coef: 0.6706 - val_loss: 0.5894 - val_dice_coef: 0.5951\n",
            "Epoch 46/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4594 - dice_coef: 0.6463 - val_loss: 0.5444 - val_dice_coef: 0.6098\n",
            "Epoch 47/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3961 - dice_coef: 0.6881 - val_loss: 0.5610 - val_dice_coef: 0.6088\n",
            "Epoch 48/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.4198 - dice_coef: 0.6923 - val_loss: 0.4892 - val_dice_coef: 0.6463\n",
            "Epoch 49/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3726 - dice_coef: 0.7248 - val_loss: 0.5124 - val_dice_coef: 0.6214\n",
            "Epoch 50/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3808 - dice_coef: 0.7192 - val_loss: 0.5161 - val_dice_coef: 0.6296\n",
            "Epoch 51/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3496 - dice_coef: 0.7441 - val_loss: 0.5104 - val_dice_coef: 0.6307\n",
            "Epoch 52/500\n",
            "34/34 [==============================] - 43s 1s/step - loss: 0.3586 - dice_coef: 0.7259 - val_loss: 0.4953 - val_dice_coef: 0.6450\n",
            "Epoch 53/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3464 - dice_coef: 0.7332 - val_loss: 0.4909 - val_dice_coef: 0.6508\n",
            "Epoch 54/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4030 - dice_coef: 0.7045 - val_loss: 0.5011 - val_dice_coef: 0.6532\n",
            "Epoch 55/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3555 - dice_coef: 0.7267 - val_loss: 0.5547 - val_dice_coef: 0.6348\n",
            "Epoch 56/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.5163 - dice_coef: 0.6343 - val_loss: 0.7346 - val_dice_coef: 0.4663\n",
            "Epoch 57/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4631 - dice_coef: 0.6526 - val_loss: 0.5138 - val_dice_coef: 0.6309\n",
            "Epoch 58/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.4551 - dice_coef: 0.6681 - val_loss: 0.5462 - val_dice_coef: 0.5942\n",
            "Epoch 59/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3892 - dice_coef: 0.7100 - val_loss: 0.5183 - val_dice_coef: 0.6303\n",
            "Epoch 60/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3536 - dice_coef: 0.7139 - val_loss: 0.4924 - val_dice_coef: 0.6348\n",
            "Epoch 61/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3220 - dice_coef: 0.7658 - val_loss: 0.4637 - val_dice_coef: 0.6643\n",
            "Epoch 62/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3251 - dice_coef: 0.7587 - val_loss: 0.5412 - val_dice_coef: 0.6286\n",
            "Epoch 63/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3115 - dice_coef: 0.7715 - val_loss: 0.4753 - val_dice_coef: 0.6765\n",
            "Epoch 64/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2897 - dice_coef: 0.7809 - val_loss: 0.5042 - val_dice_coef: 0.6543\n",
            "Epoch 65/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3198 - dice_coef: 0.7626 - val_loss: 0.4936 - val_dice_coef: 0.6533\n",
            "Epoch 66/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.3080 - dice_coef: 0.7666 - val_loss: 0.4859 - val_dice_coef: 0.6630\n",
            "Epoch 67/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2936 - dice_coef: 0.7837 - val_loss: 0.5414 - val_dice_coef: 0.6031\n",
            "Epoch 68/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2827 - dice_coef: 0.7690 - val_loss: 0.5127 - val_dice_coef: 0.6487\n",
            "Epoch 69/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2900 - dice_coef: 0.7853 - val_loss: 0.4861 - val_dice_coef: 0.6657\n",
            "Epoch 70/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2603 - dice_coef: 0.8027 - val_loss: 0.5083 - val_dice_coef: 0.6446\n",
            "Epoch 71/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2632 - dice_coef: 0.8054 - val_loss: 0.5099 - val_dice_coef: 0.6492\n",
            "Epoch 72/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2306 - dice_coef: 0.8290 - val_loss: 0.5177 - val_dice_coef: 0.6660\n",
            "Epoch 73/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2259 - dice_coef: 0.8322 - val_loss: 0.5009 - val_dice_coef: 0.6729\n",
            "Epoch 74/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2151 - dice_coef: 0.8339 - val_loss: 0.4959 - val_dice_coef: 0.6597\n",
            "Epoch 75/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2145 - dice_coef: 0.8399 - val_loss: 0.4920 - val_dice_coef: 0.6701\n",
            "Epoch 76/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2163 - dice_coef: 0.8119 - val_loss: 0.5182 - val_dice_coef: 0.6504\n",
            "Epoch 77/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2185 - dice_coef: 0.8375 - val_loss: 0.5312 - val_dice_coef: 0.6565\n",
            "Epoch 78/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1854 - dice_coef: 0.8617 - val_loss: 0.5291 - val_dice_coef: 0.6731\n",
            "Epoch 79/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1900 - dice_coef: 0.8557 - val_loss: 0.5162 - val_dice_coef: 0.6629\n",
            "Epoch 80/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1777 - dice_coef: 0.8650 - val_loss: 0.5412 - val_dice_coef: 0.6565\n",
            "Epoch 81/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1730 - dice_coef: 0.8680 - val_loss: 0.5303 - val_dice_coef: 0.6512\n",
            "Epoch 82/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1750 - dice_coef: 0.8482 - val_loss: 0.5408 - val_dice_coef: 0.6533\n",
            "Epoch 83/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2850 - dice_coef: 0.7833 - val_loss: 0.5100 - val_dice_coef: 0.6645\n",
            "Epoch 84/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2286 - dice_coef: 0.8300 - val_loss: 0.5504 - val_dice_coef: 0.6415\n",
            "Epoch 85/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1930 - dice_coef: 0.8533 - val_loss: 0.5338 - val_dice_coef: 0.6497\n",
            "Epoch 86/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1765 - dice_coef: 0.8486 - val_loss: 0.5282 - val_dice_coef: 0.6476\n",
            "Epoch 87/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1910 - dice_coef: 0.8554 - val_loss: 0.5283 - val_dice_coef: 0.6557\n",
            "Epoch 88/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1864 - dice_coef: 0.8519 - val_loss: 0.5438 - val_dice_coef: 0.6581\n",
            "Epoch 89/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2587 - dice_coef: 0.8034 - val_loss: 0.5721 - val_dice_coef: 0.6321\n",
            "Epoch 90/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.2114 - dice_coef: 0.8377 - val_loss: 0.5572 - val_dice_coef: 0.6509\n",
            "Epoch 91/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1814 - dice_coef: 0.8630 - val_loss: 0.5379 - val_dice_coef: 0.6603\n",
            "Epoch 92/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1595 - dice_coef: 0.8777 - val_loss: 0.5576 - val_dice_coef: 0.6610\n",
            "Epoch 93/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1584 - dice_coef: 0.8766 - val_loss: 0.5650 - val_dice_coef: 0.6544\n",
            "Epoch 94/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1396 - dice_coef: 0.8904 - val_loss: 0.5637 - val_dice_coef: 0.6491\n",
            "Epoch 95/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1352 - dice_coef: 0.8932 - val_loss: 0.5648 - val_dice_coef: 0.6580\n",
            "Epoch 96/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1327 - dice_coef: 0.8947 - val_loss: 0.5524 - val_dice_coef: 0.6665\n",
            "Epoch 97/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1600 - dice_coef: 0.8627 - val_loss: 0.5423 - val_dice_coef: 0.6452\n",
            "Epoch 98/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1842 - dice_coef: 0.8538 - val_loss: 0.5004 - val_dice_coef: 0.6584\n",
            "Epoch 99/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1521 - dice_coef: 0.8849 - val_loss: 0.5496 - val_dice_coef: 0.6645\n",
            "Epoch 100/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1552 - dice_coef: 0.8798 - val_loss: 0.5441 - val_dice_coef: 0.6526\n",
            "Epoch 101/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1255 - dice_coef: 0.9003 - val_loss: 0.5456 - val_dice_coef: 0.6508\n",
            "Epoch 102/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1135 - dice_coef: 0.9099 - val_loss: 0.5695 - val_dice_coef: 0.6482\n",
            "Epoch 103/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1124 - dice_coef: 0.9085 - val_loss: 0.5766 - val_dice_coef: 0.6546\n",
            "Epoch 104/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1113 - dice_coef: 0.9100 - val_loss: 0.6157 - val_dice_coef: 0.6417\n",
            "Epoch 105/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1115 - dice_coef: 0.9083 - val_loss: 0.5768 - val_dice_coef: 0.6386\n",
            "Epoch 106/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1212 - dice_coef: 0.9017 - val_loss: 0.5846 - val_dice_coef: 0.6507\n",
            "Epoch 107/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1121 - dice_coef: 0.9087 - val_loss: 0.5538 - val_dice_coef: 0.6637\n",
            "Epoch 108/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1353 - dice_coef: 0.8922 - val_loss: 0.5709 - val_dice_coef: 0.6346\n",
            "Epoch 109/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1296 - dice_coef: 0.8959 - val_loss: 0.5786 - val_dice_coef: 0.6565\n",
            "Epoch 110/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1059 - dice_coef: 0.9124 - val_loss: 0.6001 - val_dice_coef: 0.6457\n",
            "Epoch 111/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0994 - dice_coef: 0.9195 - val_loss: 0.6038 - val_dice_coef: 0.6411\n",
            "Epoch 112/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0937 - dice_coef: 0.9217 - val_loss: 0.6208 - val_dice_coef: 0.6235\n",
            "Epoch 113/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0945 - dice_coef: 0.9218 - val_loss: 0.6255 - val_dice_coef: 0.6443\n",
            "Epoch 114/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0901 - dice_coef: 0.9256 - val_loss: 0.6046 - val_dice_coef: 0.6516\n",
            "Epoch 115/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0863 - dice_coef: 0.9282 - val_loss: 0.6423 - val_dice_coef: 0.6514\n",
            "Epoch 116/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0925 - dice_coef: 0.9241 - val_loss: 0.6282 - val_dice_coef: 0.6314\n",
            "Epoch 117/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0914 - dice_coef: 0.9191 - val_loss: 0.6301 - val_dice_coef: 0.6597\n",
            "Epoch 118/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1418 - dice_coef: 0.8899 - val_loss: 0.5856 - val_dice_coef: 0.6325\n",
            "Epoch 119/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1140 - dice_coef: 0.9078 - val_loss: 0.6339 - val_dice_coef: 0.6266\n",
            "Epoch 120/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1076 - dice_coef: 0.9115 - val_loss: 0.5976 - val_dice_coef: 0.6399\n",
            "Epoch 121/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0984 - dice_coef: 0.9192 - val_loss: 0.6135 - val_dice_coef: 0.6428\n",
            "Epoch 122/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0903 - dice_coef: 0.9247 - val_loss: 0.6021 - val_dice_coef: 0.6488\n",
            "Epoch 123/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0827 - dice_coef: 0.9307 - val_loss: 0.6485 - val_dice_coef: 0.6475\n",
            "Epoch 124/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0773 - dice_coef: 0.9333 - val_loss: 0.6963 - val_dice_coef: 0.6369\n",
            "Epoch 125/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0754 - dice_coef: 0.9355 - val_loss: 0.6466 - val_dice_coef: 0.6508\n",
            "Epoch 126/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0724 - dice_coef: 0.9375 - val_loss: 0.6172 - val_dice_coef: 0.6501\n",
            "Epoch 127/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0754 - dice_coef: 0.9325 - val_loss: 0.6858 - val_dice_coef: 0.6344\n",
            "Epoch 128/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0927 - dice_coef: 0.9237 - val_loss: 0.6483 - val_dice_coef: 0.6451\n",
            "Epoch 129/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0821 - dice_coef: 0.9295 - val_loss: 0.5983 - val_dice_coef: 0.6539\n",
            "Epoch 130/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0791 - dice_coef: 0.9319 - val_loss: 0.6832 - val_dice_coef: 0.6344\n",
            "Epoch 131/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0756 - dice_coef: 0.9363 - val_loss: 0.6392 - val_dice_coef: 0.6543\n",
            "Epoch 132/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0677 - dice_coef: 0.9422 - val_loss: 0.6554 - val_dice_coef: 0.6507\n",
            "Epoch 133/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0661 - dice_coef: 0.9408 - val_loss: 0.6450 - val_dice_coef: 0.6546\n",
            "Epoch 134/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0712 - dice_coef: 0.9372 - val_loss: 0.6686 - val_dice_coef: 0.6525\n",
            "Epoch 135/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0722 - dice_coef: 0.9397 - val_loss: 0.6580 - val_dice_coef: 0.6504\n",
            "Epoch 136/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0688 - dice_coef: 0.9416 - val_loss: 0.6745 - val_dice_coef: 0.6556\n",
            "Epoch 137/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0615 - dice_coef: 0.9446 - val_loss: 0.6806 - val_dice_coef: 0.6387\n",
            "Epoch 138/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0825 - dice_coef: 0.9307 - val_loss: 0.6542 - val_dice_coef: 0.6500\n",
            "Epoch 139/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0830 - dice_coef: 0.9289 - val_loss: 0.6089 - val_dice_coef: 0.6452\n",
            "Epoch 140/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1032 - dice_coef: 0.9154 - val_loss: 0.6551 - val_dice_coef: 0.6297\n",
            "Epoch 141/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0844 - dice_coef: 0.9295 - val_loss: 0.6023 - val_dice_coef: 0.6555\n",
            "Epoch 142/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0712 - dice_coef: 0.9387 - val_loss: 0.6281 - val_dice_coef: 0.6572\n",
            "Epoch 143/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0741 - dice_coef: 0.9380 - val_loss: 0.6739 - val_dice_coef: 0.6353\n",
            "Epoch 144/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0662 - dice_coef: 0.9393 - val_loss: 0.7102 - val_dice_coef: 0.6424\n",
            "Epoch 145/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1455 - dice_coef: 0.8878 - val_loss: 0.6888 - val_dice_coef: 0.6321\n",
            "Epoch 146/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1613 - dice_coef: 0.8790 - val_loss: 0.6455 - val_dice_coef: 0.6078\n",
            "Epoch 147/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1066 - dice_coef: 0.9129 - val_loss: 0.6060 - val_dice_coef: 0.6449\n",
            "Epoch 148/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.1010 - dice_coef: 0.9167 - val_loss: 0.6871 - val_dice_coef: 0.6399\n",
            "Epoch 149/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0899 - dice_coef: 0.9258 - val_loss: 0.6338 - val_dice_coef: 0.6551\n",
            "Epoch 150/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0731 - dice_coef: 0.9382 - val_loss: 0.6640 - val_dice_coef: 0.6478\n",
            "Epoch 151/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0656 - dice_coef: 0.9429 - val_loss: 0.6717 - val_dice_coef: 0.6456\n",
            "Epoch 152/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0632 - dice_coef: 0.9437 - val_loss: 0.7051 - val_dice_coef: 0.6328\n",
            "Epoch 153/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0615 - dice_coef: 0.9463 - val_loss: 0.7209 - val_dice_coef: 0.6410\n",
            "Epoch 154/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0577 - dice_coef: 0.9485 - val_loss: 0.7004 - val_dice_coef: 0.6441\n",
            "Epoch 155/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0560 - dice_coef: 0.9506 - val_loss: 0.7187 - val_dice_coef: 0.6456\n",
            "Epoch 156/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0523 - dice_coef: 0.9531 - val_loss: 0.7097 - val_dice_coef: 0.6478\n",
            "Epoch 157/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0548 - dice_coef: 0.9516 - val_loss: 0.7968 - val_dice_coef: 0.6344\n",
            "Epoch 158/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0550 - dice_coef: 0.9530 - val_loss: 0.6653 - val_dice_coef: 0.6414\n",
            "Epoch 159/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0500 - dice_coef: 0.9552 - val_loss: 0.7543 - val_dice_coef: 0.6430\n",
            "Epoch 160/500\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0515 - dice_coef: 0.9545 - val_loss: 0.6973 - val_dice_coef: 0.6508\n",
            "Epoch 161/500\n",
            "34/34 [==============================] - ETA: 0s - loss: 0.0498 - dice_coef: 0.9557Restoring model weights from the end of the best epoch: 61.\n",
            "34/34 [==============================] - 42s 1s/step - loss: 0.0498 - dice_coef: 0.9557 - val_loss: 0.7189 - val_dice_coef: 0.6535\n",
            "Epoch 161: early stopping\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model.save('/content/gdrive/My Drive/BreastUSDatasetNPYs/Combined_UNet_Model.h5')"
      ],
      "metadata": {
        "id": "btIWU3ijiD4t"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Make predictions on test set\n",
        "testPreds = model.predict(combinedTestImages)\n",
        "print(testPreds.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0yPJeKV-7OuL",
        "outputId": "76b2208d-6c12-4aae-aac0-96e13f0a987c"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(166, 224, 224, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate the test set predictions based on Dice Score\n",
        "dScoreList = list()\n",
        "\n",
        "for i in range(len(testPreds)):\n",
        "  yTrue = combinedTestMasks[i].reshape((224,224))\n",
        "  yPred = testPreds[i].reshape((224,224))\n",
        "  yTrue = tf.convert_to_tensor(yTrue, np.float32)\n",
        "  yTrue = K.cast(yTrue, 'float32')\n",
        "  yPred = tf.convert_to_tensor(yPred, np.float32)\n",
        "  yPred = K.cast(yPred, 'float32')\n",
        "  dScoreList.append(float(1-dice_loss(yTrue, yPred)))\n",
        "\n",
        "print(dScoreList)\n",
        "print('The average Dice Score of the test set segmentation is: ',np.mean(dScoreList))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zd0tQeO87ZCj",
        "outputId": "ce962965-d70d-4cdd-808f-b096db814153"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0.09170430898666382, 0.845727801322937, 0.7470941543579102, 0.970902144908905, 0.01332634687423706, 0.1370307207107544, 0.7913942933082581, 0.09965306520462036, 0.2106572389602661, 0.8720808029174805, 0.7620186805725098, 0.5638324022293091, 0.362601637840271, 0.6197812557220459, 0.2423211932182312, 0.9351905584335327, 0.9329071044921875, 0.9312455058097839, 0.9096270203590393, 0.9351567625999451, 0.2175004482269287, 0.4959160089492798, 0.9230114221572876, 0.6633973121643066, 0.6507247686386108, 0.42320185899734497, 0.5620959401130676, 0.9043064713478088, 0.915550708770752, 0.9261789321899414, 0.1721305251121521, 0.20695841312408447, 0.9724377989768982, 0.8711205720901489, 0.9321907758712769, 0.6474726796150208, 0.9378888010978699, 0.8896368741989136, 0.9554059505462646, 0.9413787126541138, 0.13902318477630615, 0.8946936130523682, 0.8590953350067139, 0.9315415024757385, 0.8995092511177063, 0.5887740254402161, 0.9010825157165527, 0.8483723998069763, 0.6634516716003418, 0.6235185861587524, 0.43006372451782227, 0.9472151398658752, 0.9278643727302551, 0.3104248046875, 0.42002570629119873, 0.8527671694755554, 0.9128527045249939, 0.46259093284606934, 0.6590029001235962, 0.859344482421875, 0.9764300584793091, 0.921721875667572, 0.6763272285461426, 0.8795953989028931, 0.8791590332984924, 0.915946364402771, 0.9302532076835632, 0.1631629467010498, 0.8601882457733154, 0.49809473752975464, 0.9494218826293945, 0.8955039978027344, 0.2804105281829834, 0.8460327982902527, 0.38308095932006836, 0.7275648713111877, 0.5030529499053955, 0.23592376708984375, 0.8911260962486267, 0.3939216136932373, 0.8539260625839233, 0.5198462605476379, 0.900789737701416, 0.8593804836273193, 0.7142640352249146, 0.5257203578948975, 0.5746418237686157, 0.517036497592926, 0.6881831288337708, 0.9123461246490479, 0.9053082466125488, 0.7652916312217712, 0.881580114364624, 0.12790250778198242, 0.6544874310493469, 0.9034039974212646, 0.7248291373252869, 0.6787562966346741, 0.04277753829956055, 0.8733371496200562, 0.8878206610679626, 0.9091463685035706, 0.7370266914367676, 0.8043843507766724, 0.9361979961395264, 0.8930992484092712, 0.5382044315338135, 0.94905686378479, 0.8430938720703125, 0.40794479846954346, 0.02128756046295166, 0.9160783886909485, 0.6026856303215027, 0.06300246715545654, 0.9245572090148926, 0.8252304792404175, 0.06888473033905029, 0.1945420503616333, 0.8213140368461609, 0.4112069010734558, 0.22831547260284424, 0.9569883346557617, 0.8403476476669312, 0.9403903484344482, 0.15153419971466064, 0.9270579218864441, 0.831282913684845, 0.8800493478775024, 0.6825966835021973, 0.6736220717430115, 0.8134666681289673, 0.9261633157730103, 0.8375067710876465, 0.5180067420005798, 0.03527885675430298, 0.926476001739502, 0.8920127153396606, 0.7910435199737549, 0.5568834543228149, 0.345708429813385, 0.5952029228210449, 0.9110392332077026, 0.9345871210098267, 0.862862765789032, 0.8780795335769653, 0.9293500781059265, 0.7688162326812744, 0.8485072255134583, 0.890180230140686, 0.8815974593162537, 0.913760781288147, 0.45762836933135986, 0.6532865166664124, 0.8244398832321167, 0.8602176308631897, 0.3344306945800781, 0.925080418586731, 0.5450384020805359, 0.9606357216835022, 0.9562341570854187, 0.937341570854187, 0.7522616982460022, 0.815328061580658, 0.9501408338546753, 0.9258291125297546, 0.8510215878486633]\n",
            "The average Dice Score of the test set segmentation is:  0.6952836958758802\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "We can observe that the automatic segmentation model is able to achieve a decent average Dice Score of ~70% on the combined test set."
      ],
      "metadata": {
        "id": "dal-ct3e8pw6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Display a few sample predictions: the best 3 and the worst 3\n",
        "\n",
        "indList = [i for i in range(len(combinedTestImages))]\n",
        "\n",
        "sortedInds = [x for _,x in sorted(zip(dScoreList,indList))]\n",
        "\n",
        "# Get the best and the worst scoring predictions\n",
        "\n",
        "bestInds = [sortedInds[-1],sortedInds[-2],sortedInds[-3]]\n",
        "worstInds = [sortedInds[0],sortedInds[1],sortedInds[2]]\n",
        "\n",
        "# Display the 3 best predictions\n",
        "\n",
        "print('The following are the best test predictions in the order: Original BUS, G.T. Mask, Prediction')\n",
        "fig, axs = plt.subplots(3, 3)\n",
        "axs[0, 0].imshow(combinedTestImages[bestInds[0]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[0, 1].imshow(combinedTestMasks[bestInds[0]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[0, 2].imshow(testPreds[bestInds[0]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[1, 0].imshow(combinedTestImages[bestInds[1]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[1, 1].imshow(combinedTestMasks[bestInds[1]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[1, 2].imshow(testPreds[bestInds[1]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[2, 0].imshow(combinedTestImages[bestInds[2]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[2, 1].imshow(combinedTestMasks[bestInds[2]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[2, 2].imshow(testPreds[bestInds[2]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "# remove the x and y ticks\n",
        "for ax in axs.flat:\n",
        "    ax.set_xticks([])\n",
        "    ax.set_yticks([])\n",
        "\n",
        "# Hide x labels and tick labels for top plots and y ticks for right plots.\n",
        "for ax in axs.flat:\n",
        "    ax.label_outer()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "id": "sJ8AJJ9c821z",
        "outputId": "99c4f4ca-931e-4523-87a7-95c57c20dc82"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The following are the best test predictions in the order: Original BUS, G.T. Mask, Prediction\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 9 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Display the 3 worst predictions\n",
        "\n",
        "print('The following are the worst test predictions in the order: Original BUS, G.T. Mask, Prediction')\n",
        "fig, axs = plt.subplots(3, 3)\n",
        "axs[0, 0].imshow(combinedTestImages[worstInds[0]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[0, 1].imshow(combinedTestMasks[worstInds[0]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[0, 2].imshow(testPreds[worstInds[0]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[1, 0].imshow(combinedTestImages[worstInds[1]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[1, 1].imshow(combinedTestMasks[worstInds[1]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[1, 2].imshow(testPreds[worstInds[1]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[2, 0].imshow(combinedTestImages[worstInds[2]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[2, 1].imshow(combinedTestMasks[worstInds[2]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "axs[2, 2].imshow(testPreds[worstInds[2]].reshape((224,224)), cmap='gray')\n",
        "\n",
        "# remove the x and y ticks\n",
        "for ax in axs.flat:\n",
        "    ax.set_xticks([])\n",
        "    ax.set_yticks([])\n",
        "\n",
        "# Hide x labels and tick labels for top plots and y ticks for right plots.\n",
        "for ax in axs.flat:\n",
        "    ax.label_outer()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "id": "rUIM6Z-a_bRl",
        "outputId": "753ea9fd-485b-4195-a7ce-c5de9bf010b9"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The following are the worst test predictions in the order: Original BUS, G.T. Mask, Prediction\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 9 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    }
  ]
}