InitialImplementation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 5,
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.8.8"
    },
    "colab": {
      "name": "Untitled.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "379d0a54-3026-41f8-85a0-367fd77424f6",
        "outputId": "720b8bd7-4ae0-4141-c5f5-ac7e37fa5da4"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "import numpy as np\n",
        "from glob import glob\n",
        "import matplotlib.pyplot as plt\n",
        "import cv2\n",
        "import os\n",
        "from google.colab import drive\n",
        "print(tf.__version__)\n",
        "\n",
        "from tensorflow.keras import regularizers\n",
        "from tensorflow.keras.layers import Input, Conv2D, BatchNormalization, Activation, MaxPool2D, UpSampling2D, Concatenate, Conv2DTranspose\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping\n",
        "from tensorflow.keras.metrics import Recall, Precision\n",
        "from tensorflow.keras import backend as K"
      ],
      "id": "379d0a54-3026-41f8-85a0-367fd77424f6",
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2.5.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9kJHLkai79Hd",
        "outputId": "042383e0-8054-4136-d6ac-af1ae1c2173a"
      },
      "source": [
        "drive.mount('/content/drive')"
      ],
      "id": "9kJHLkai79Hd",
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fPBs0dy98Kup",
        "outputId": "49bbbcda-3603-4c36-adf3-cfde9a20cbcf"
      },
      "source": [
        "%cd drive/MyDrive/Edge/\n",
        "!ls"
      ],
      "id": "fPBs0dy98Kup",
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/MyDrive/Edge\n",
            "BIPED  DexiNed.h5\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dedd5e4a-a4ec-4efc-9e7c-abf94fec1bb8"
      },
      "source": [
        "train_path = \"./BIPED/edges/imgs/train/rgbr/real/*\"\n",
        "edge_train_path = \"./BIPED/edges/edge_maps/train/rgbr/real/*\"\n",
        "test_path = \"./BIPED/edges/imgs/test/rgbr/*\"\n",
        "edge_test_path = \"./BIPED/edges/edge_maps/test/rgbr/*\""
      ],
      "id": "dedd5e4a-a4ec-4efc-9e7c-abf94fec1bb8",
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "e0e1c946-a32b-4821-a35d-9b2d6261fdaf"
      },
      "source": [
        "def load_data(ipath, epath):\n",
        "    images = sorted(glob(os.path.join(ipath)))\n",
        "    edges = sorted(glob(os.path.join(epath)))\n",
        "    return images, edges"
      ],
      "id": "e0e1c946-a32b-4821-a35d-9b2d6261fdaf",
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8b4aa7de-03b4-4568-b18c-b658089cbaf4"
      },
      "source": [
        "images, edges = load_data(train_path, edge_train_path)\n",
        "testimg, testedg = load_data(test_path, edge_test_path)"
      ],
      "id": "8b4aa7de-03b4-4568-b18c-b658089cbaf4",
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8accce3a-24b1-4783-86cf-260c88646b73",
        "outputId": "953ef58b-a426-4ede-e8b5-eb83d36481ae"
      },
      "source": [
        "print(len(images), len(edges))"
      ],
      "id": "8accce3a-24b1-4783-86cf-260c88646b73",
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "200 200\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c3bbd9d5-4ae7-43a9-9694-2d9fe28fc65b"
      },
      "source": [
        "def read_image(path, H=720, W=1280):\n",
        "    x = cv2.imread(path, cv2.IMREAD_COLOR)\n",
        "    x = cv2.resize(x, (W,H))\n",
        "    x = x/255.0\n",
        "    x = x.astype(np.float32)\n",
        "    return x\n",
        "\n",
        "def read_edge(path, H=720, W=1280):\n",
        "    x = cv2.imread(path, cv2.IMREAD_GRAYSCALE)\n",
        "    _, x = cv2.threshold(x,127,255,cv2.THRESH_BINARY)\n",
        "    x = cv2.resize(x, (W,H))\n",
        "    x = np.expand_dims(x, axis=-1)\n",
        "    x = x/255.0\n",
        "    x = x.astype(np.float32)\n",
        "    return x"
      ],
      "id": "c3bbd9d5-4ae7-43a9-9694-2d9fe28fc65b",
      "execution_count": 49,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "f83e1219-8534-4996-83b4-989f5f068090"
      },
      "source": [
        "def preprocess(x,y,H=720, W=1280):\n",
        "    \n",
        "    def f(x,y):\n",
        "        x = x.decode()\n",
        "        y = y.decode()\n",
        "        \n",
        "        x = read_image(x)\n",
        "        y = read_edge(y)\n",
        "        return x, y\n",
        "    \n",
        "    images, edges = tf.numpy_function(f, [x, y], [tf.float32, tf.float32])\n",
        "    images.set_shape([H, W, 3])\n",
        "    edges.set_shape([H, W, 1])\n",
        "    return images, edges"
      ],
      "id": "f83e1219-8534-4996-83b4-989f5f068090",
      "execution_count": 50,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "30c1fd06-77ca-4de5-be6f-48d3be5867a3"
      },
      "source": [
        "def tf_data(x,y):\n",
        "    data = tf.data.Dataset.from_tensor_slices((x,y))\n",
        "    data = data.shuffle(buffer_size=10)\n",
        "    data = data.map(preprocess)\n",
        "    data = data.batch(1)\n",
        "    data = data.prefetch(1)\n",
        "    return data"
      ],
      "id": "30c1fd06-77ca-4de5-be6f-48d3be5867a3",
      "execution_count": 51,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7e3120eb-0e37-475d-b3fc-668eff4012ba"
      },
      "source": [
        "train_data = tf_data(images, edges)\n",
        "test_data = tf_data(testimg, testedg)"
      ],
      "id": "7e3120eb-0e37-475d-b3fc-668eff4012ba",
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6ba732b9-0046-4181-b353-488620a5b033",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "48650b73-b7ee-4b08-8378-49aac564ad8e"
      },
      "source": [
        "for x, y in train_data:\n",
        "    print(x.shape, y.shape)\n",
        "print(train_data)"
      ],
      "id": "6ba732b9-0046-4181-b353-488620a5b033",
      "execution_count": 53,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "(1, 720, 1280, 3) (1, 720, 1280, 1)\n",
            "<PrefetchDataset shapes: ((None, 720, 1280, 3), (None, 720, 1280, 1)), types: (tf.float32, tf.float32)>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "e33230bb-053a-4733-93ce-263ad3d8b809"
      },
      "source": [
        "weight_init = tf.initializers.glorot_uniform()\n",
        "\n",
        "l2 = regularizers.l2\n",
        "w_decay=1e-3"
      ],
      "id": "e33230bb-053a-4733-93ce-263ad3d8b809",
      "execution_count": 54,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "888d5e5b-e521-4fc3-995d-e3290f5388db"
      },
      "source": [
        "def double_conv_block(inputs, f1, f2, stride=(1,1), act=True):\n",
        "    \n",
        "    k_reg = None if w_decay is None else l2(w_decay)\n",
        "    \n",
        "    c1 = Conv2D(filters=f1, kernel_size=(3,3), strides=stride, padding=\"same\", kernel_initializer=weight_init, kernel_regularizer=k_reg)(inputs)\n",
        "    c1 = BatchNormalization()(c1)\n",
        "    c1 = Conv2D(filters=f2, kernel_size=(3,3), strides=(1,1), padding=\"same\", kernel_initializer=weight_init, kernel_regularizer=k_reg)(c1)\n",
        "    c1 = BatchNormalization()(c1)\n",
        "    \n",
        "    if act:\n",
        "        c1 = Activation(\"relu\")(c1)\n",
        "    \n",
        "    return c1"
      ],
      "id": "888d5e5b-e521-4fc3-995d-e3290f5388db",
      "execution_count": 55,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "523e6c13-7430-4a36-b949-f8e617b0f25b"
      },
      "source": [
        "def DenseBlock(inputs, filters):\n",
        "    \n",
        "    k_reg = None if w_decay is None else l2(w_decay)\n",
        "    \n",
        "    c1 = Activation(\"relu\")(inputs)\n",
        "    c1 = Conv2D(filters=filters, kernel_size=(3,3), strides=(1,1), padding='same', kernel_initializer=weight_init, kernel_regularizer=k_reg)(c1)\n",
        "    c1 = BatchNormalization()(c1)\n",
        "    c1 = Activation(\"relu\")(c1)\n",
        "    c1 = Conv2D(filters=filters, kernel_size=(3,3), strides=(1,1), padding='same', kernel_initializer=weight_init, kernel_regularizer=k_reg)(c1)\n",
        "    c1 = BatchNormalization()(c1)\n",
        "    \n",
        "    return c1"
      ],
      "id": "523e6c13-7430-4a36-b949-f8e617b0f25b",
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d6d8ca3e-3ef6-47ee-85ad-2a4f03f33cca"
      },
      "source": [
        "def single_conv_block(inputs, filters, kernel=(1,1), strides=(1,1), bn=False, act=False, w_init=None):\n",
        "    \n",
        "    k_reg = None if w_decay is None else l2(w_decay)\n",
        "    \n",
        "    c1 = Conv2D(filters=filters, kernel_size=kernel, strides=strides, padding=\"same\", kernel_initializer=w_init, kernel_regularizer=k_reg)(inputs)\n",
        "    \n",
        "    if bn:\n",
        "        c1 = BatchNormalization()(c1)\n",
        "        \n",
        "    if act:\n",
        "        c1 = Activation(\"relu\")(c1)\n",
        "        \n",
        "    return c1"
      ],
      "id": "d6d8ca3e-3ef6-47ee-85ad-2a4f03f33cca",
      "execution_count": 57,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "768fe149-63a1-4997-8a4d-e636a6af5d60"
      },
      "source": [
        "def UpConvBlock(inputs, up_scale):\n",
        "    \n",
        "    constant_features = 16\n",
        "    k_reg = None if w_decay is None else l2(w_decay)\n",
        "    c1 = inputs\n",
        "    \n",
        "    total_up_scale = 2 ** up_scale\n",
        "    \n",
        "    for i in range(up_scale):\n",
        "        \n",
        "        filters = 1 if i == up_scale-1 else constant_features\n",
        "        \n",
        "        if i==up_scale-1:\n",
        "            \n",
        "            c1 = Conv2D(filters=filters, kernel_size=(1,1), \n",
        "                        strides=(1,1), padding='same', activation='relu', \n",
        "                        kernel_initializer=tf.initializers.TruncatedNormal(mean=0.),kernel_regularizer=k_reg)(c1)\n",
        "            c1 = Conv2DTranspose(filters, kernel_size=(total_up_scale,total_up_scale), \n",
        "                                 strides=(2,2), padding='same', kernel_initializer=tf.initializers.TruncatedNormal(stddev=0.1), \n",
        "                                 kernel_regularizer=k_reg)(c1)\n",
        "        else:\n",
        "            \n",
        "            c1 = Conv2D(filters=filters, kernel_size=(1,1), \n",
        "                        strides=(1,1), padding='same', activation='relu', \n",
        "                        kernel_initializer=tf.initializers.TruncatedNormal(mean=0.),kernel_regularizer=k_reg)(c1)\n",
        "            c1 = Conv2DTranspose(filters, kernel_size=(total_up_scale,total_up_scale), strides=(2,2), padding='same',\n",
        "                                kernel_initializer=weight_init, kernel_regularizer=k_reg)(c1)\n",
        "            \n",
        "    \n",
        "    \n",
        "    return c1   "
      ],
      "id": "768fe149-63a1-4997-8a4d-e636a6af5d60",
      "execution_count": 58,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "df6d7e48-37cd-4b5f-af22-84807b741417"
      },
      "source": [
        "def build_model():\n",
        "    \n",
        "    inputs = Input((720, 1280,3))\n",
        "    \n",
        "    b1 = double_conv_block(inputs, 32, 64, stride=(2,2), act=False)\n",
        "    side_b1 = single_conv_block(b1, 128, strides=(2,2), bn=True, w_init=weight_init)\n",
        "    \n",
        "    b2 = double_conv_block(b1, 128, 128, act=False)\n",
        "    b2 = MaxPool2D(pool_size=(3, 3), strides=2, padding='same')(b2)\n",
        "    b2_up = Concatenate()([b2 + side_b1])\n",
        "    side_b2 = single_conv_block(b2_up, 256, strides=(2,2), bn=True, w_init=weight_init)\n",
        "    \n",
        "    b3_pre = single_conv_block(b2, 128, strides=(1,1), bn=True, w_init=weight_init)\n",
        "    b3 = DenseBlock(b2_up+b3_pre, 256)\n",
        "    b3 = DenseBlock(b3, 256)\n",
        "    b3_down = MaxPool2D(pool_size=(3,3), strides=2, padding=\"same\")(b3)\n",
        "    b3_add = b3_down + side_b2\n",
        "    side_b3 = single_conv_block(b3_add, 512, strides=(2,2), bn=True, w_init=weight_init)\n",
        "    \n",
        "    b4_pre_1 = single_conv_block(b2, 256, strides=(2,2), w_init=weight_init)\n",
        "    b4_pre_2 = single_conv_block(b3_down+b4_pre_1, 256, strides=(1,1), bn=True, w_init=weight_init)\n",
        "    b4 = DenseBlock(b3_add+b4_pre_2, 512)\n",
        "    b4 = DenseBlock(b4, 512)\n",
        "    b4 = DenseBlock(b4, 512)\n",
        "    b4_down = MaxPool2D(pool_size=(3,3), strides=(2,2), padding=\"same\")(b4)\n",
        "    b4_add = b4_down+side_b3\n",
        "    side_b4 = single_conv_block(b4_add, 512, strides=(1,1), bn=True, w_init=weight_init)\n",
        "    \n",
        "    b5_pre_1 = single_conv_block(b4_pre_1, 512, strides=(2,2), w_init=weight_init)\n",
        "    b5_pre_2 = single_conv_block(b4_down+b5_pre_1, 512, strides=(1,1), bn=True, w_init=weight_init)\n",
        "    b5 = DenseBlock(b4_add+b5_pre_2, 512)\n",
        "    b5 = DenseBlock(b5, 512)\n",
        "    b5 = DenseBlock(b5, 512)\n",
        "    b5_add = b5+side_b4\n",
        "    \n",
        "    b6_pre = single_conv_block(b5, 512, strides=(1,1), bn=True, w_init=weight_init)\n",
        "    b6 = DenseBlock(b5_add+b6_pre, 256)\n",
        "    b6 = DenseBlock(b6, 256)\n",
        "    b6 = DenseBlock(b6, 256)\n",
        "    \n",
        "    out1 = UpConvBlock(b1, 1)\n",
        "    out2 = UpConvBlock(b2, 2)\n",
        "    out3 = UpConvBlock(b3, 2)\n",
        "    out4 = UpConvBlock(b4, 3)\n",
        "    out5 = UpConvBlock(b5, 4)\n",
        "    out6 = UpConvBlock(b6, 4)\n",
        "    results = [out1, out2, out3, out4, out5, out6]\n",
        "    \n",
        "    block_cat = Concatenate()([out1,out2,out3,out4,out5,out6])\n",
        "    b1_cat = single_conv_block(block_cat, 3, strides=(1,1), bn=True, w_init=tf.constant_initializer(1/5))\n",
        "    bl = Concatenate()([block_cat, b1_cat])\n",
        "    bl = Conv2D(1, kernel_size=(3,3), padding=\"same\")(bl)\n",
        "    bl = Activation(\"sigmoid\")(bl)\n",
        "\n",
        "    return Model(inputs=inputs, outputs=bl)"
      ],
      "id": "df6d7e48-37cd-4b5f-af22-84807b741417",
      "execution_count": 59,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "59882019-e068-487e-961e-c2580015d2cd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "72725c40-30b9-4621-b93b-16cbce9e33d4"
      },
      "source": [
        "model = build_model()\n",
        "model.summary()"
      ],
      "id": "59882019-e068-487e-961e-c2580015d2cd",
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 720, 1280, 3 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d (Conv2D)                 (None, 360, 640, 32) 896         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization (BatchNorma (None, 360, 640, 32) 128         conv2d[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_1 (Conv2D)               (None, 360, 640, 64) 18496       batch_normalization[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_1 (BatchNor (None, 360, 640, 64) 256         conv2d_1[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_3 (Conv2D)               (None, 360, 640, 128 73856       batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_3 (BatchNor (None, 360, 640, 128 512         conv2d_3[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_4 (Conv2D)               (None, 360, 640, 128 147584      batch_normalization_3[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_4 (BatchNor (None, 360, 640, 128 512         conv2d_4[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_2 (Conv2D)               (None, 180, 320, 128 8320        batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D)    (None, 180, 320, 128 0           batch_normalization_4[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_2 (BatchNor (None, 180, 320, 128 512         conv2d_2[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add (TFOpLambd (None, 180, 320, 128 0           max_pooling2d[0][0]              \n",
            "                                                                 batch_normalization_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_6 (Conv2D)               (None, 180, 320, 128 16512       max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "concatenate (Concatenate)       (None, 180, 320, 128 0           tf.__operators__.add[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_6 (BatchNor (None, 180, 320, 128 512         conv2d_6[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_1 (TFOpLam (None, 180, 320, 128 0           concatenate[0][0]                \n",
            "                                                                 batch_normalization_6[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "activation (Activation)         (None, 180, 320, 128 0           tf.__operators__.add_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_7 (Conv2D)               (None, 180, 320, 256 295168      activation[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_7 (BatchNor (None, 180, 320, 256 1024        conv2d_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_1 (Activation)       (None, 180, 320, 256 0           batch_normalization_7[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_8 (Conv2D)               (None, 180, 320, 256 590080      activation_1[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_8 (BatchNor (None, 180, 320, 256 1024        conv2d_8[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_2 (Activation)       (None, 180, 320, 256 0           batch_normalization_8[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_9 (Conv2D)               (None, 180, 320, 256 590080      activation_2[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_9 (BatchNor (None, 180, 320, 256 1024        conv2d_9[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "activation_3 (Activation)       (None, 180, 320, 256 0           batch_normalization_9[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_10 (Conv2D)              (None, 180, 320, 256 590080      activation_3[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_10 (BatchNo (None, 180, 320, 256 1024        conv2d_10[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2D)  (None, 90, 160, 256) 0           batch_normalization_10[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_12 (Conv2D)              (None, 90, 160, 256) 33024       max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_5 (Conv2D)               (None, 90, 160, 256) 33024       concatenate[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_3 (TFOpLam (None, 90, 160, 256) 0           max_pooling2d_1[0][0]            \n",
            "                                                                 conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_5 (BatchNor (None, 90, 160, 256) 1024        conv2d_5[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_13 (Conv2D)              (None, 90, 160, 256) 65792       tf.__operators__.add_3[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_2 (TFOpLam (None, 90, 160, 256) 0           max_pooling2d_1[0][0]            \n",
            "                                                                 batch_normalization_5[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_12 (BatchNo (None, 90, 160, 256) 1024        conv2d_13[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_4 (TFOpLam (None, 90, 160, 256) 0           tf.__operators__.add_2[0][0]     \n",
            "                                                                 batch_normalization_12[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_4 (Activation)       (None, 90, 160, 256) 0           tf.__operators__.add_4[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_14 (Conv2D)              (None, 90, 160, 512) 1180160     activation_4[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_13 (BatchNo (None, 90, 160, 512) 2048        conv2d_14[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_5 (Activation)       (None, 90, 160, 512) 0           batch_normalization_13[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_15 (Conv2D)              (None, 90, 160, 512) 2359808     activation_5[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_14 (BatchNo (None, 90, 160, 512) 2048        conv2d_15[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_6 (Activation)       (None, 90, 160, 512) 0           batch_normalization_14[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_16 (Conv2D)              (None, 90, 160, 512) 2359808     activation_6[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_15 (BatchNo (None, 90, 160, 512) 2048        conv2d_16[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_7 (Activation)       (None, 90, 160, 512) 0           batch_normalization_15[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_17 (Conv2D)              (None, 90, 160, 512) 2359808     activation_7[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_16 (BatchNo (None, 90, 160, 512) 2048        conv2d_17[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_8 (Activation)       (None, 90, 160, 512) 0           batch_normalization_16[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_18 (Conv2D)              (None, 90, 160, 512) 2359808     activation_8[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_17 (BatchNo (None, 90, 160, 512) 2048        conv2d_18[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_9 (Activation)       (None, 90, 160, 512) 0           batch_normalization_17[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_19 (Conv2D)              (None, 90, 160, 512) 2359808     activation_9[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_18 (BatchNo (None, 90, 160, 512) 2048        conv2d_19[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2D)  (None, 45, 80, 512)  0           batch_normalization_18[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_21 (Conv2D)              (None, 45, 80, 512)  131584      conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_11 (Conv2D)              (None, 45, 80, 512)  131584      tf.__operators__.add_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_6 (TFOpLam (None, 45, 80, 512)  0           max_pooling2d_2[0][0]            \n",
            "                                                                 conv2d_21[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_11 (BatchNo (None, 45, 80, 512)  2048        conv2d_11[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_22 (Conv2D)              (None, 45, 80, 512)  262656      tf.__operators__.add_6[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_5 (TFOpLam (None, 45, 80, 512)  0           max_pooling2d_2[0][0]            \n",
            "                                                                 batch_normalization_11[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_20 (BatchNo (None, 45, 80, 512)  2048        conv2d_22[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_7 (TFOpLam (None, 45, 80, 512)  0           tf.__operators__.add_5[0][0]     \n",
            "                                                                 batch_normalization_20[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_10 (Activation)      (None, 45, 80, 512)  0           tf.__operators__.add_7[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_23 (Conv2D)              (None, 45, 80, 512)  2359808     activation_10[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_21 (BatchNo (None, 45, 80, 512)  2048        conv2d_23[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_11 (Activation)      (None, 45, 80, 512)  0           batch_normalization_21[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_24 (Conv2D)              (None, 45, 80, 512)  2359808     activation_11[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_22 (BatchNo (None, 45, 80, 512)  2048        conv2d_24[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_12 (Activation)      (None, 45, 80, 512)  0           batch_normalization_22[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_25 (Conv2D)              (None, 45, 80, 512)  2359808     activation_12[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_23 (BatchNo (None, 45, 80, 512)  2048        conv2d_25[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_13 (Activation)      (None, 45, 80, 512)  0           batch_normalization_23[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_26 (Conv2D)              (None, 45, 80, 512)  2359808     activation_13[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_24 (BatchNo (None, 45, 80, 512)  2048        conv2d_26[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_14 (Activation)      (None, 45, 80, 512)  0           batch_normalization_24[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_27 (Conv2D)              (None, 45, 80, 512)  2359808     activation_14[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_25 (BatchNo (None, 45, 80, 512)  2048        conv2d_27[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_15 (Activation)      (None, 45, 80, 512)  0           batch_normalization_25[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_28 (Conv2D)              (None, 45, 80, 512)  2359808     activation_15[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_26 (BatchNo (None, 45, 80, 512)  2048        conv2d_28[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_20 (Conv2D)              (None, 45, 80, 512)  262656      tf.__operators__.add_5[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_19 (BatchNo (None, 45, 80, 512)  2048        conv2d_20[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_29 (Conv2D)              (None, 45, 80, 512)  262656      batch_normalization_26[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_8 (TFOpLam (None, 45, 80, 512)  0           batch_normalization_26[0][0]     \n",
            "                                                                 batch_normalization_19[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_27 (BatchNo (None, 45, 80, 512)  2048        conv2d_29[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_9 (TFOpLam (None, 45, 80, 512)  0           tf.__operators__.add_8[0][0]     \n",
            "                                                                 batch_normalization_27[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "activation_16 (Activation)      (None, 45, 80, 512)  0           tf.__operators__.add_9[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_30 (Conv2D)              (None, 45, 80, 256)  1179904     activation_16[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_28 (BatchNo (None, 45, 80, 256)  1024        conv2d_30[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_17 (Activation)      (None, 45, 80, 256)  0           batch_normalization_28[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_31 (Conv2D)              (None, 45, 80, 256)  590080      activation_17[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_29 (BatchNo (None, 45, 80, 256)  1024        conv2d_31[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_18 (Activation)      (None, 45, 80, 256)  0           batch_normalization_29[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_32 (Conv2D)              (None, 45, 80, 256)  590080      activation_18[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_30 (BatchNo (None, 45, 80, 256)  1024        conv2d_32[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_19 (Activation)      (None, 45, 80, 256)  0           batch_normalization_30[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_33 (Conv2D)              (None, 45, 80, 256)  590080      activation_19[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_31 (BatchNo (None, 45, 80, 256)  1024        conv2d_33[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_20 (Activation)      (None, 45, 80, 256)  0           batch_normalization_31[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_34 (Conv2D)              (None, 45, 80, 256)  590080      activation_20[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_32 (BatchNo (None, 45, 80, 256)  1024        conv2d_34[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "activation_21 (Activation)      (None, 45, 80, 256)  0           batch_normalization_32[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_35 (Conv2D)              (None, 45, 80, 256)  590080      activation_21[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_33 (BatchNo (None, 45, 80, 256)  1024        conv2d_35[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_44 (Conv2D)              (None, 45, 80, 16)   8208        batch_normalization_26[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_48 (Conv2D)              (None, 45, 80, 16)   4112        batch_normalization_33[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_8 (Conv2DTrans (None, 90, 160, 16)  65552       conv2d_44[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_12 (Conv2DTran (None, 90, 160, 16)  65552       conv2d_48[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_41 (Conv2D)              (None, 90, 160, 16)  8208        batch_normalization_18[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_45 (Conv2D)              (None, 90, 160, 16)  272         conv2d_transpose_8[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_49 (Conv2D)              (None, 90, 160, 16)  272         conv2d_transpose_12[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_5 (Conv2DTrans (None, 180, 320, 16) 16400       conv2d_41[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_9 (Conv2DTrans (None, 180, 320, 16) 65552       conv2d_45[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_13 (Conv2DTran (None, 180, 320, 16) 65552       conv2d_49[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_37 (Conv2D)              (None, 180, 320, 16) 2064        max_pooling2d[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_39 (Conv2D)              (None, 180, 320, 16) 4112        batch_normalization_10[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_42 (Conv2D)              (None, 180, 320, 16) 272         conv2d_transpose_5[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_46 (Conv2D)              (None, 180, 320, 16) 272         conv2d_transpose_9[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_50 (Conv2D)              (None, 180, 320, 16) 272         conv2d_transpose_13[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_1 (Conv2DTrans (None, 360, 640, 16) 4112        conv2d_37[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_3 (Conv2DTrans (None, 360, 640, 16) 4112        conv2d_39[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_6 (Conv2DTrans (None, 360, 640, 16) 16400       conv2d_42[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_10 (Conv2DTran (None, 360, 640, 16) 65552       conv2d_46[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_14 (Conv2DTran (None, 360, 640, 16) 65552       conv2d_50[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_36 (Conv2D)              (None, 360, 640, 1)  65          batch_normalization_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_38 (Conv2D)              (None, 360, 640, 1)  17          conv2d_transpose_1[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_40 (Conv2D)              (None, 360, 640, 1)  17          conv2d_transpose_3[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_43 (Conv2D)              (None, 360, 640, 1)  17          conv2d_transpose_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_47 (Conv2D)              (None, 360, 640, 1)  17          conv2d_transpose_10[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_51 (Conv2D)              (None, 360, 640, 1)  17          conv2d_transpose_14[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose (Conv2DTranspo (None, 720, 1280, 1) 5           conv2d_36[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_2 (Conv2DTrans (None, 720, 1280, 1) 17          conv2d_38[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_4 (Conv2DTrans (None, 720, 1280, 1) 17          conv2d_40[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_7 (Conv2DTrans (None, 720, 1280, 1) 65          conv2d_43[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_11 (Conv2DTran (None, 720, 1280, 1) 257         conv2d_47[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_transpose_15 (Conv2DTran (None, 720, 1280, 1) 257         conv2d_51[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_1 (Concatenate)     (None, 720, 1280, 6) 0           conv2d_transpose[0][0]           \n",
            "                                                                 conv2d_transpose_2[0][0]         \n",
            "                                                                 conv2d_transpose_4[0][0]         \n",
            "                                                                 conv2d_transpose_7[0][0]         \n",
            "                                                                 conv2d_transpose_11[0][0]        \n",
            "                                                                 conv2d_transpose_15[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_52 (Conv2D)              (None, 720, 1280, 3) 21          concatenate_1[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_34 (BatchNo (None, 720, 1280, 3) 12          conv2d_52[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_2 (Concatenate)     (None, 720, 1280, 9) 0           concatenate_1[0][0]              \n",
            "                                                                 batch_normalization_34[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_53 (Conv2D)              (None, 720, 1280, 1) 82          concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "activation_22 (Activation)      (None, 720, 1280, 1) 0           conv2d_53[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 35,293,171\n",
            "Trainable params: 35,269,421\n",
            "Non-trainable params: 23,750\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "73b8ee3a-3e03-4679-b4ea-71da211bd1be"
      },
      "source": [
        "model.compile(\n",
        "    loss=\"binary_crossentropy\", \n",
        "    optimizer=\"adam\", \n",
        "    metrics=[tf.keras.metrics.BinaryAccuracy(), \n",
        "             tf.keras.metrics.FalseNegatives(), \n",
        "             tf.keras.metrics.FalsePositives(), \n",
        "             tf.keras.metrics.TruePositives(), \n",
        "             tf.keras.metrics.TrueNegatives()])"
      ],
      "id": "73b8ee3a-3e03-4679-b4ea-71da211bd1be",
      "execution_count": 63,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "655646ce-3b24-4fb6-b816-31f1a128af72",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6c75b54c-4f71-49d2-e3f4-4650af40b92b"
      },
      "source": [
        "train_steps=len(train_data)\n",
        "val_steps=len(test_data)\n",
        "print(train_steps)\n",
        "model.fit(train_data, validation_data=test_data, epochs=20, steps_per_epoch=train_steps, validation_steps=val_steps)"
      ],
      "id": "655646ce-3b24-4fb6-b816-31f1a128af72",
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "200\n",
            "Epoch 1/20\n",
            "200/200 [==============================] - 146s 697ms/step - loss: 0.1815 - binary_accuracy: 0.9697 - false_negatives_1: 5318489.0000 - false_positives_1: 274404.0000 - true_positives_1: 92265.0000 - true_negatives_1: 178634848.0000 - val_loss: 0.2168 - val_binary_accuracy: 0.9674 - val_false_negatives_1: 1501545.0000 - val_false_positives_1: 14.0000 - val_true_positives_1: 10.0000 - val_true_negatives_1: 44578440.0000\n",
            "Epoch 2/20\n",
            "200/200 [==============================] - 137s 687ms/step - loss: 0.1534 - binary_accuracy: 0.9696 - false_negatives_1: 5331454.0000 - false_positives_1: 272649.0000 - true_positives_1: 79300.0000 - true_negatives_1: 178636560.0000 - val_loss: 0.1473 - val_binary_accuracy: 0.9673 - val_false_negatives_1: 1500514.0000 - val_false_positives_1: 5715.0000 - val_true_positives_1: 1041.0000 - val_true_negatives_1: 44572732.0000\n",
            "Epoch 3/20\n",
            "200/200 [==============================] - 138s 688ms/step - loss: 0.1355 - binary_accuracy: 0.9696 - false_negatives_1: 5339284.0000 - false_positives_1: 265519.0000 - true_positives_1: 71470.0000 - true_negatives_1: 178643744.0000 - val_loss: 0.1512 - val_binary_accuracy: 0.9672 - val_false_negatives_1: 1495644.0000 - val_false_positives_1: 16370.0000 - val_true_positives_1: 5911.0000 - val_true_negatives_1: 44562080.0000\n",
            "Epoch 4/20\n",
            "200/200 [==============================] - 137s 688ms/step - loss: 0.1274 - binary_accuracy: 0.9696 - false_negatives_1: 5331637.0000 - false_positives_1: 276718.0000 - true_positives_1: 79117.0000 - true_negatives_1: 178632576.0000 - val_loss: 0.1378 - val_binary_accuracy: 0.9672 - val_false_negatives_1: 1494882.0000 - val_false_positives_1: 17394.0000 - val_true_positives_1: 6673.0000 - val_true_negatives_1: 44561048.0000\n",
            "Epoch 5/20\n",
            "200/200 [==============================] - 137s 687ms/step - loss: 0.1256 - binary_accuracy: 0.9696 - false_negatives_1: 5333805.0000 - false_positives_1: 266333.0000 - true_positives_1: 76949.0000 - true_negatives_1: 178642976.0000 - val_loss: 0.1370 - val_binary_accuracy: 0.9673 - val_false_negatives_1: 1497612.0000 - val_false_positives_1: 9542.0000 - val_true_positives_1: 3943.0000 - val_true_negatives_1: 44568896.0000\n",
            "Epoch 6/20\n",
            "200/200 [==============================] - 136s 681ms/step - loss: 0.1226 - binary_accuracy: 0.9696 - false_negatives_1: 5336746.0000 - false_positives_1: 263136.0000 - true_positives_1: 74008.0000 - true_negatives_1: 178646112.0000 - val_loss: 0.1917 - val_binary_accuracy: 0.9616 - val_false_negatives_1: 1367860.0000 - val_false_positives_1: 402925.0000 - val_true_positives_1: 133695.0000 - val_true_negatives_1: 44175524.0000\n",
            "Epoch 7/20\n",
            "200/200 [==============================] - 130s 649ms/step - loss: 0.1228 - binary_accuracy: 0.9696 - false_negatives_1: 5336573.0000 - false_positives_1: 261919.0000 - true_positives_1: 74181.0000 - true_negatives_1: 178647360.0000 - val_loss: 0.1349 - val_binary_accuracy: 0.9671 - val_false_negatives_1: 1490677.0000 - val_false_positives_1: 23319.0000 - val_true_positives_1: 10878.0000 - val_true_negatives_1: 44555124.0000\n",
            "Epoch 8/20\n",
            "200/200 [==============================] - 127s 636ms/step - loss: 0.1217 - binary_accuracy: 0.9696 - false_negatives_1: 5330183.0000 - false_positives_1: 277343.0000 - true_positives_1: 80571.0000 - true_negatives_1: 178631872.0000 - val_loss: 0.1477 - val_binary_accuracy: 0.9670 - val_false_negatives_1: 1483498.0000 - val_false_positives_1: 39384.0000 - val_true_positives_1: 18057.0000 - val_true_negatives_1: 44539068.0000\n",
            "Epoch 9/20\n",
            "200/200 [==============================] - 127s 633ms/step - loss: 0.1210 - binary_accuracy: 0.9696 - false_negatives_1: 5334616.0000 - false_positives_1: 260659.0000 - true_positives_1: 76138.0000 - true_negatives_1: 178648672.0000 - val_loss: 0.1341 - val_binary_accuracy: 0.9671 - val_false_negatives_1: 1489366.0000 - val_false_positives_1: 25764.0000 - val_true_positives_1: 12189.0000 - val_true_negatives_1: 44552676.0000\n",
            "Epoch 10/20\n",
            "200/200 [==============================] - 127s 633ms/step - loss: 0.1209 - binary_accuracy: 0.9697 - false_negatives_1: 5333750.0000 - false_positives_1: 258999.0000 - true_positives_1: 77004.0000 - true_negatives_1: 178650240.0000 - val_loss: 0.1374 - val_binary_accuracy: 0.9672 - val_false_negatives_1: 1494605.0000 - val_false_positives_1: 16683.0000 - val_true_positives_1: 6950.0000 - val_true_negatives_1: 44561776.0000\n",
            "Epoch 11/20\n",
            "200/200 [==============================] - 127s 633ms/step - loss: 0.1206 - binary_accuracy: 0.9697 - false_negatives_1: 5332225.0000 - false_positives_1: 255164.0000 - true_positives_1: 78529.0000 - true_negatives_1: 178654048.0000 - val_loss: 0.1357 - val_binary_accuracy: 0.9664 - val_false_negatives_1: 1481872.0000 - val_false_positives_1: 66629.0000 - val_true_positives_1: 19683.0000 - val_true_negatives_1: 44511816.0000\n",
            "Epoch 12/20\n",
            "200/200 [==============================] - 126s 629ms/step - loss: 0.1208 - binary_accuracy: 0.9696 - false_negatives_1: 5331832.0000 - false_positives_1: 264094.0000 - true_positives_1: 78922.0000 - true_negatives_1: 178645120.0000 - val_loss: 0.1349 - val_binary_accuracy: 0.9673 - val_false_negatives_1: 1496519.0000 - val_false_positives_1: 10562.0000 - val_true_positives_1: 5036.0000 - val_true_negatives_1: 44567884.0000\n",
            "Epoch 13/20\n",
            "200/200 [==============================] - 126s 628ms/step - loss: 0.1207 - binary_accuracy: 0.9696 - false_negatives_1: 5333870.0000 - false_positives_1: 261677.0000 - true_positives_1: 76884.0000 - true_negatives_1: 178647536.0000 - val_loss: 0.1522 - val_binary_accuracy: 0.9674 - val_false_negatives_1: 1499481.0000 - val_false_positives_1: 4607.0000 - val_true_positives_1: 2074.0000 - val_true_negatives_1: 44573840.0000\n",
            "Epoch 14/20\n",
            "200/200 [==============================] - 126s 631ms/step - loss: 0.1207 - binary_accuracy: 0.9697 - false_negatives_1: 5333838.0000 - false_positives_1: 259568.0000 - true_positives_1: 76916.0000 - true_negatives_1: 178649728.0000 - val_loss: 0.1400 - val_binary_accuracy: 0.9649 - val_false_negatives_1: 1428511.0000 - val_false_positives_1: 190134.0000 - val_true_positives_1: 73044.0000 - val_true_negatives_1: 44388304.0000\n",
            "Epoch 15/20\n",
            "200/200 [==============================] - 126s 630ms/step - loss: 0.1206 - binary_accuracy: 0.9696 - false_negatives_1: 5334731.0000 - false_positives_1: 262219.0000 - true_positives_1: 76023.0000 - true_negatives_1: 178647056.0000 - val_loss: 0.1702 - val_binary_accuracy: 0.9585 - val_false_negatives_1: 1371550.0000 - val_false_positives_1: 538735.0000 - val_true_positives_1: 130005.0000 - val_true_negatives_1: 44039708.0000\n",
            "Epoch 16/20\n",
            "200/200 [==============================] - 126s 629ms/step - loss: 0.1207 - binary_accuracy: 0.9696 - false_negatives_1: 5333546.0000 - false_positives_1: 265797.0000 - true_positives_1: 77208.0000 - true_negatives_1: 178643408.0000 - val_loss: 0.1342 - val_binary_accuracy: 0.9673 - val_false_negatives_1: 1495580.0000 - val_false_positives_1: 12567.0000 - val_true_positives_1: 5975.0000 - val_true_negatives_1: 44565876.0000\n",
            "Epoch 17/20\n",
            "200/200 [==============================] - 126s 629ms/step - loss: 0.1205 - binary_accuracy: 0.9696 - false_negatives_1: 5330890.0000 - false_positives_1: 266266.0000 - true_positives_1: 79864.0000 - true_negatives_1: 178643040.0000 - val_loss: 0.1394 - val_binary_accuracy: 0.9651 - val_false_negatives_1: 1432289.0000 - val_false_positives_1: 174193.0000 - val_true_positives_1: 69266.0000 - val_true_negatives_1: 44404264.0000\n",
            "Epoch 18/20\n",
            "200/200 [==============================] - 126s 629ms/step - loss: 0.1207 - binary_accuracy: 0.9696 - false_negatives_1: 5333597.0000 - false_positives_1: 263973.0000 - true_positives_1: 77157.0000 - true_negatives_1: 178645344.0000 - val_loss: 0.1727 - val_binary_accuracy: 0.9637 - val_false_negatives_1: 1396218.0000 - val_false_positives_1: 278388.0000 - val_true_positives_1: 105337.0000 - val_true_negatives_1: 44300056.0000\n",
            "Epoch 19/20\n",
            "200/200 [==============================] - 126s 630ms/step - loss: 0.1204 - binary_accuracy: 0.9696 - false_negatives_1: 5331932.0000 - false_positives_1: 271547.0000 - true_positives_1: 78822.0000 - true_negatives_1: 178637776.0000 - val_loss: 0.1308 - val_binary_accuracy: 0.9671 - val_false_negatives_1: 1493516.0000 - val_false_positives_1: 20990.0000 - val_true_positives_1: 8039.0000 - val_true_negatives_1: 44557464.0000\n",
            "Epoch 20/20\n",
            "200/200 [==============================] - 126s 628ms/step - loss: 0.1207 - binary_accuracy: 0.9697 - false_negatives_1: 5336475.0000 - false_positives_1: 253738.0000 - true_positives_1: 74279.0000 - true_negatives_1: 178655472.0000 - val_loss: 0.1537 - val_binary_accuracy: 0.9638 - val_false_negatives_1: 1403322.0000 - val_false_positives_1: 264593.0000 - val_true_positives_1: 98233.0000 - val_true_negatives_1: 44313856.0000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f4daedc3610>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 64
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "47948f01-dda4-44f3-88ec-958db390b3a2"
      },
      "source": [
        "model.save('./DexiNed.h5')"
      ],
      "id": "47948f01-dda4-44f3-88ec-958db390b3a2",
      "execution_count": 65,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ynBvCPwFJ5nt"
      },
      "source": [
        "img = read_image('./BIPED/edges/edge_maps/test/rgbr/RGB_008.png')\n",
        "img = np.expand_dims(img, axis=0)"
      ],
      "id": "ynBvCPwFJ5nt",
      "execution_count": 66,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d7rOHFJ1KJEq"
      },
      "source": [
        "res = model.predict(img)"
      ],
      "id": "d7rOHFJ1KJEq",
      "execution_count": 74,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kIL32ciDK68k",
        "outputId": "1e1980b7-70d5-4a37-e24b-ab7f8a8d2d5d"
      },
      "source": [
        "print(res.shape)\n",
        "res = res[-1,:,:]\n",
        "print(res.shape)"
      ],
      "id": "kIL32ciDK68k",
      "execution_count": 75,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1, 720, 1280, 1)\n",
            "(720, 1280, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-O_9GA9rhKtW",
        "outputId": "c4187cd7-4653-4d79-c191-bf91cc63dc6e"
      },
      "source": [
        "res = res[:,:,-1]\n",
        "print(res.shape)"
      ],
      "id": "-O_9GA9rhKtW",
      "execution_count": 76,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(720, 1280)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 240
        },
        "id": "4jYS-TdYLyTP",
        "outputId": "65598c66-94fd-44da-a347-811f115f0751"
      },
      "source": [
        "plt.imshow(res, cmap=\"gray\")\n",
        "plt.show()"
      ],
      "id": "4jYS-TdYLyTP",
      "execution_count": 77,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 257
        },
        "id": "YnbMKuCkNdMG",
        "outputId": "776afb43-88c3-48f1-e23c-49a43fc122b3"
      },
      "source": [
        "img = read_image('./BIPED/edges/edge_maps/test/rgbr/RGB_008.png')\n",
        "plt.imshow(img)"
      ],
      "id": "YnbMKuCkNdMG",
      "execution_count": 78,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f4da0166090>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 78
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_ectW5AWOqJg"
      },
      "source": [
        ""
      ],
      "id": "_ectW5AWOqJg",
      "execution_count": null,
      "outputs": []
    }
  ]
}