MIRNet_v2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "L03da8Lvd0Gr"
      },
      "outputs": [],
      "source": [
        "import os\n",
        "import cv2\n",
        "import random\n",
        "import numpy as np\n",
        "from glob import glob\n",
        "from PIL import Image, ImageOps\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive"
      ],
      "metadata": {
        "id": "ZI19ptKef2yi"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install --upgrade --no-cache-dir gdown"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "i3wx_uhCgtJO",
        "outputId": "00c59743-cf8d-4f9b-e8a0-cbc93c2cf436"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: gdown in /usr/local/lib/python3.8/dist-packages (4.4.0)\n",
            "Collecting gdown\n",
            "  Downloading gdown-4.6.0-py3-none-any.whl (14 kB)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.8/dist-packages (from gdown) (4.64.1)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.8/dist-packages (from gdown) (1.15.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.8/dist-packages (from gdown) (3.9.0)\n",
            "Requirement already satisfied: requests[socks] in /usr/local/lib/python3.8/dist-packages (from gdown) (2.25.1)\n",
            "Requirement already satisfied: beautifulsoup4 in /usr/local/lib/python3.8/dist-packages (from gdown) (4.6.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.8/dist-packages (from requests[socks]->gdown) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.8/dist-packages (from requests[socks]->gdown) (2022.12.7)\n",
            "Requirement already satisfied: chardet<5,>=3.0.2 in /usr/local/lib/python3.8/dist-packages (from requests[socks]->gdown) (4.0.0)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.8/dist-packages (from requests[socks]->gdown) (1.24.3)\n",
            "Requirement already satisfied: PySocks!=1.5.7,>=1.5.6 in /usr/local/lib/python3.8/dist-packages (from requests[socks]->gdown) (1.7.1)\n",
            "Installing collected packages: gdown\n",
            "  Attempting uninstall: gdown\n",
            "    Found existing installation: gdown 4.4.0\n",
            "    Uninstalling gdown-4.4.0:\n",
            "      Successfully uninstalled gdown-4.4.0\n",
            "Successfully installed gdown-4.6.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!gdown https://drive.google.com/uc?id=1DdGIJ4PZPlF2ikl8mNM9V-PdVxVLbQi6\n",
        "!unzip -q lol_dataset.zip"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AbdSGamseazQ",
        "outputId": "b2777abd-41fd-4088-a31c-03651f34d938"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1DdGIJ4PZPlF2ikl8mNM9V-PdVxVLbQi6\n",
            "To: /content/lol_dataset.zip\n",
            "100% 347M/347M [00:03<00:00, 113MB/s]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "random.seed(10)\n",
        "\n",
        "IMAGE_SIZE = 128\n",
        "BATCH_SIZE = 4\n",
        "MAX_TRAIN_IMAGES = 300\n",
        "\n",
        "\n",
        "def read_image(image_path):\n",
        "    image = tf.io.read_file(image_path)\n",
        "    image = tf.image.decode_png(image, channels=3)\n",
        "    image.set_shape([None, None, 3])\n",
        "    image = tf.cast(image, dtype=tf.float32) / 255.0\n",
        "    return image\n",
        "\n",
        "\n",
        "def random_crop(low_image, enhanced_image):\n",
        "    low_image_shape = tf.shape(low_image)[:2]\n",
        "    low_w = tf.random.uniform(\n",
        "        shape=(), maxval=low_image_shape[1] - IMAGE_SIZE + 1, dtype=tf.int32\n",
        "    )\n",
        "    low_h = tf.random.uniform(\n",
        "        shape=(), maxval=low_image_shape[0] - IMAGE_SIZE + 1, dtype=tf.int32\n",
        "    )\n",
        "    enhanced_w = low_w\n",
        "    enhanced_h = low_h\n",
        "    low_image_cropped = low_image[\n",
        "        low_h : low_h + IMAGE_SIZE, low_w : low_w + IMAGE_SIZE\n",
        "    ]\n",
        "    enhanced_image_cropped = enhanced_image[\n",
        "        enhanced_h : enhanced_h + IMAGE_SIZE, enhanced_w : enhanced_w + IMAGE_SIZE\n",
        "    ]\n",
        "    return low_image_cropped, enhanced_image_cropped\n",
        "\n",
        "\n",
        "def load_data(low_light_image_path, enhanced_image_path):\n",
        "    low_light_image = read_image(low_light_image_path)\n",
        "    enhanced_image = read_image(enhanced_image_path)\n",
        "    low_light_image, enhanced_image = random_crop(low_light_image, enhanced_image)\n",
        "    return low_light_image, enhanced_image\n",
        "\n",
        "\n",
        "def get_dataset(low_light_images, enhanced_images):\n",
        "    dataset = tf.data.Dataset.from_tensor_slices((low_light_images, enhanced_images))\n",
        "    dataset = dataset.map(load_data, num_parallel_calls=tf.data.AUTOTUNE)\n",
        "    dataset = dataset.batch(BATCH_SIZE, drop_remainder=True)\n",
        "    return dataset\n",
        "\n",
        "\n",
        "train_low_light_images = sorted(glob(\"./lol_dataset/our485/low/*\"))[:MAX_TRAIN_IMAGES]\n",
        "train_enhanced_images = sorted(glob(\"./lol_dataset/our485/high/*\"))[:MAX_TRAIN_IMAGES]\n",
        "\n",
        "val_low_light_images = sorted(glob(\"./lol_dataset/our485/low/*\"))[MAX_TRAIN_IMAGES:]\n",
        "val_enhanced_images = sorted(glob(\"./lol_dataset/our485/high/*\"))[MAX_TRAIN_IMAGES:]\n",
        "\n",
        "test_low_light_images = sorted(glob(\"./lol_dataset/eval15/low/*\"))\n",
        "test_enhanced_images = sorted(glob(\"./lol_dataset/eval15/high/*\"))\n",
        "\n",
        "\n",
        "train_dataset = get_dataset(train_low_light_images, train_enhanced_images)\n",
        "val_dataset = get_dataset(val_low_light_images, val_enhanced_images)\n",
        "\n",
        "\n",
        "print(\"Train Dataset:\", train_dataset)\n",
        "print(\"Val Dataset:\", val_dataset)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PKrOOY_hebge",
        "outputId": "af519278-e6d2-4ed1-e5e5-b7c20b2984f7"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Train Dataset: <BatchDataset element_spec=(TensorSpec(shape=(4, None, None, 3), dtype=tf.float32, name=None), TensorSpec(shape=(4, None, None, 3), dtype=tf.float32, name=None))>\n",
            "Val Dataset: <BatchDataset element_spec=(TensorSpec(shape=(4, None, None, 3), dtype=tf.float32, name=None), TensorSpec(shape=(4, None, None, 3), dtype=tf.float32, name=None))>\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def selective_kernel_feature_fusion(\n",
        "    multi_scale_feature_1, multi_scale_feature_2, multi_scale_feature_3\n",
        "):\n",
        "    channels = list(multi_scale_feature_1.shape)[-1]\n",
        "    combined_feature = layers.Add()(\n",
        "        [multi_scale_feature_1, multi_scale_feature_2, multi_scale_feature_3]\n",
        "    )\n",
        "    gap = layers.GlobalAveragePooling2D()(combined_feature)\n",
        "    channel_wise_statistics = tf.reshape(gap, shape=(-1, 1, 1, channels))\n",
        "    compact_feature_representation = layers.Conv2D(\n",
        "        filters=channels // 8, kernel_size=(1, 1), activation=\"relu\"\n",
        "    )(channel_wise_statistics)\n",
        "    feature_descriptor_1 = layers.Conv2D(\n",
        "        channels, kernel_size=(1, 1), activation=\"softmax\"\n",
        "    )(compact_feature_representation)\n",
        "    feature_descriptor_2 = layers.Conv2D(\n",
        "        channels, kernel_size=(1, 1), activation=\"softmax\"\n",
        "    )(compact_feature_representation)\n",
        "    feature_descriptor_3 = layers.Conv2D(\n",
        "        channels, kernel_size=(1, 1), activation=\"softmax\"\n",
        "    )(compact_feature_representation)\n",
        "    feature_1 = multi_scale_feature_1 * feature_descriptor_1\n",
        "    feature_2 = multi_scale_feature_2 * feature_descriptor_2\n",
        "    feature_3 = multi_scale_feature_3 * feature_descriptor_3\n",
        "    aggregated_feature = layers.Add()([feature_1, feature_2, feature_3])\n",
        "    return aggregated_feature"
      ],
      "metadata": {
        "id": "fNZkOFpRegV_"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def spatial_attention_block(input_tensor):\n",
        "    average_pooling = tf.reduce_max(input_tensor, axis=-1)\n",
        "    average_pooling = tf.expand_dims(average_pooling, axis=-1)\n",
        "    max_pooling = tf.reduce_mean(input_tensor, axis=-1)\n",
        "    max_pooling = tf.expand_dims(max_pooling, axis=-1)\n",
        "    concatenated = layers.Concatenate(axis=-1)([average_pooling, max_pooling])\n",
        "    feature_map = layers.Conv2D(1, kernel_size=(1, 1))(concatenated)\n",
        "    feature_map = tf.nn.sigmoid(feature_map)\n",
        "    return input_tensor * feature_map\n",
        "\n",
        "\n",
        "def channel_attention_block(input_tensor):\n",
        "    channels = list(input_tensor.shape)[-1]\n",
        "    average_pooling = layers.GlobalAveragePooling2D()(input_tensor)\n",
        "    feature_descriptor = tf.reshape(average_pooling, shape=(-1, 1, 1, channels))\n",
        "    feature_activations = layers.Conv2D(\n",
        "        filters=channels // 8, kernel_size=(1, 1), activation=\"relu\"\n",
        "    )(feature_descriptor)\n",
        "    feature_activations = layers.Conv2D(\n",
        "        filters=channels, kernel_size=(1, 1), activation=\"sigmoid\"\n",
        "    )(feature_activations)\n",
        "    return input_tensor * feature_activations\n",
        "\n",
        "\n",
        "def dual_attention_unit_block(input_tensor):\n",
        "    channels = list(input_tensor.shape)[-1]\n",
        "    feature_map = layers.Conv2D(\n",
        "        channels, kernel_size=(3, 3), padding=\"same\", activation=\"relu\"\n",
        "    )(input_tensor)\n",
        "    feature_map = layers.Conv2D(channels, kernel_size=(3, 3), padding=\"same\")(\n",
        "        feature_map\n",
        "    )\n",
        "    channel_attention = channel_attention_block(feature_map)\n",
        "    spatial_attention = spatial_attention_block(feature_map)\n",
        "    concatenation = layers.Concatenate(axis=-1)([channel_attention, spatial_attention])\n",
        "    concatenation = layers.Conv2D(channels, kernel_size=(1, 1))(concatenation)\n",
        "    return layers.Add()([input_tensor, concatenation])"
      ],
      "metadata": {
        "id": "Ms1M7LtWekIm"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Recursive Residual Modules\n",
        "\n",
        "\n",
        "def down_sampling_module(input_tensor):\n",
        "    channels = list(input_tensor.shape)[-1]\n",
        "    main_branch = layers.Conv2D(channels, kernel_size=(1, 1), activation=\"relu\")(\n",
        "        input_tensor\n",
        "    )\n",
        "    main_branch = layers.Conv2D(\n",
        "        channels, kernel_size=(3, 3), padding=\"same\", activation=\"relu\"\n",
        "    )(main_branch)\n",
        "    main_branch = layers.MaxPooling2D()(main_branch)\n",
        "    main_branch = layers.Conv2D(channels * 2, kernel_size=(1, 1))(main_branch)\n",
        "    skip_branch = layers.MaxPooling2D()(input_tensor)\n",
        "    skip_branch = layers.Conv2D(channels * 2, kernel_size=(1, 1))(skip_branch)\n",
        "    return layers.Add()([skip_branch, main_branch])\n",
        "\n",
        "\n",
        "def up_sampling_module(input_tensor):\n",
        "    channels = list(input_tensor.shape)[-1]\n",
        "    main_branch = layers.Conv2D(channels, kernel_size=(1, 1), activation=\"relu\")(\n",
        "        input_tensor\n",
        "    )\n",
        "    main_branch = layers.Conv2D(\n",
        "        channels, kernel_size=(3, 3), padding=\"same\", activation=\"relu\"\n",
        "    )(main_branch)\n",
        "    main_branch = layers.UpSampling2D()(main_branch)\n",
        "    main_branch = layers.Conv2D(channels // 2, kernel_size=(1, 1))(main_branch)\n",
        "    skip_branch = layers.UpSampling2D()(input_tensor)\n",
        "    skip_branch = layers.Conv2D(channels // 2, kernel_size=(1, 1))(skip_branch)\n",
        "    return layers.Add()([skip_branch, main_branch])\n",
        "\n",
        "\n",
        "# MRB Block\n",
        "def multi_scale_residual_block(input_tensor, channels):\n",
        "    # features\n",
        "    level1 = input_tensor\n",
        "    level2 = down_sampling_module(input_tensor)\n",
        "    level3 = down_sampling_module(level2)\n",
        "    # DAU\n",
        "    level1_dau = dual_attention_unit_block(level1)\n",
        "    level2_dau = dual_attention_unit_block(level2)\n",
        "    level3_dau = dual_attention_unit_block(level3)\n",
        "    # SKFF\n",
        "    level1_skff = selective_kernel_feature_fusion(\n",
        "        level1_dau,\n",
        "        up_sampling_module(level2_dau),\n",
        "        up_sampling_module(up_sampling_module(level3_dau)),\n",
        "    )\n",
        "    level2_skff = selective_kernel_feature_fusion(\n",
        "        down_sampling_module(level1_dau), level2_dau, up_sampling_module(level3_dau)\n",
        "    )\n",
        "    level3_skff = selective_kernel_feature_fusion(\n",
        "        down_sampling_module(down_sampling_module(level1_dau)),\n",
        "        down_sampling_module(level2_dau),\n",
        "        level3_dau,\n",
        "    )\n",
        "    # DAU 2\n",
        "    level1_dau_2 = dual_attention_unit_block(level1_skff)\n",
        "    level2_dau_2 = up_sampling_module((dual_attention_unit_block(level2_skff)))\n",
        "    level3_dau_2 = up_sampling_module(\n",
        "        up_sampling_module(dual_attention_unit_block(level3_skff))\n",
        "    )\n",
        "    # SKFF 2\n",
        "    skff_ = selective_kernel_feature_fusion(level1_dau_2, level2_dau_2, level3_dau_2)\n",
        "    conv = layers.Conv2D(channels, kernel_size=(3, 3), padding=\"same\")(skff_)\n",
        "    return layers.Add()([input_tensor, conv])"
      ],
      "metadata": {
        "id": "6WyT3teeeqA0"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def recursive_residual_group(input_tensor, num_mrb, channels):\n",
        "    conv1 = layers.Conv2D(channels, kernel_size=(3, 3), padding=\"same\")(input_tensor)\n",
        "    for _ in range(num_mrb):\n",
        "        conv1 = multi_scale_residual_block(conv1, channels)\n",
        "    conv2 = layers.Conv2D(channels, kernel_size=(3, 3), padding=\"same\")(conv1)\n",
        "    return layers.Add()([conv2, input_tensor])\n",
        "\n",
        "\n",
        "def mirnet_model(num_rrg, num_mrb, channels):\n",
        "    input_tensor = keras.Input(shape=[None, None, 3])\n",
        "    x1 = layers.Conv2D(channels, kernel_size=(3, 3), padding=\"same\")(input_tensor)\n",
        "    for _ in range(num_rrg):\n",
        "        x1 = recursive_residual_group(x1, num_mrb, channels)\n",
        "    conv = layers.Conv2D(3, kernel_size=(3, 3), padding=\"same\")(x1)\n",
        "    output_tensor = layers.Add()([input_tensor, conv])\n",
        "    return keras.Model(input_tensor, output_tensor)\n",
        "\n",
        "\n",
        "model = mirnet_model(num_rrg=2, num_mrb=2, channels=64)"
      ],
      "metadata": {
        "id": "mzqR-BeMevzY"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def charbonnier_loss(y_true, y_pred):\n",
        "    return tf.reduce_mean(tf.sqrt(tf.square(y_true - y_pred) + tf.square(1e-3)))\n",
        "\n",
        "\n",
        "def peak_signal_noise_ratio(y_true, y_pred):\n",
        "    return tf.image.psnr(y_pred, y_true, max_val=255.0)\n",
        "\n",
        "\n",
        "optimizer = keras.optimizers.Adam(learning_rate=1e-4)\n",
        "model.compile(\n",
        "    optimizer=optimizer, loss=charbonnier_loss, metrics=[peak_signal_noise_ratio]\n",
        ")\n",
        "\n",
        "history = model.fit(\n",
        "    train_dataset,\n",
        "    validation_data=val_dataset,\n",
        "    epochs=50,\n",
        "    callbacks=[\n",
        "        keras.callbacks.ReduceLROnPlateau(\n",
        "            monitor=\"val_peak_signal_noise_ratio\",\n",
        "            factor=0.5,\n",
        "            patience=5,\n",
        "            verbose=1,\n",
        "            min_delta=1e-7,\n",
        "            mode=\"max\",\n",
        "        )\n",
        "    ],\n",
        ")\n",
        "\n",
        "plt.plot(history.history[\"loss\"], label=\"train_loss\")\n",
        "plt.plot(history.history[\"val_loss\"], label=\"val_loss\")\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.title(\"Train and Validation Losses Over Epochs\", fontsize=14)\n",
        "plt.legend()\n",
        "plt.grid()\n",
        "plt.show()\n",
        "\n",
        "\n",
        "plt.plot(history.history[\"peak_signal_noise_ratio\"], label=\"train_psnr\")\n",
        "plt.plot(history.history[\"val_peak_signal_noise_ratio\"], label=\"val_psnr\")\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"PSNR\")\n",
        "plt.title(\"Train and Validation PSNR Over Epochs\", fontsize=14)\n",
        "plt.legend()\n",
        "plt.grid()\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "m9JJ7LAve0oX",
        "outputId": "5e0b0d97-6fdf-4d2f-d330-c55953552d0a"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n",
            "75/75 [==============================] - 121s 1s/step - loss: 0.2198 - peak_signal_noise_ratio: 61.6170 - val_loss: 0.1313 - val_peak_signal_noise_ratio: 65.6103 - lr: 1.0000e-04\n",
            "Epoch 2/50\n",
            "75/75 [==============================] - 80s 1s/step - loss: 0.1718 - peak_signal_noise_ratio: 63.5055 - val_loss: 0.1279 - val_peak_signal_noise_ratio: 65.6145 - lr: 1.0000e-04\n",
            "Epoch 3/50\n",
            "75/75 [==============================] - 71s 945ms/step - loss: 0.1624 - peak_signal_noise_ratio: 63.8900 - val_loss: 0.1317 - val_peak_signal_noise_ratio: 65.1716 - lr: 1.0000e-04\n",
            "Epoch 4/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1633 - peak_signal_noise_ratio: 63.8619 - val_loss: 0.1164 - val_peak_signal_noise_ratio: 66.3549 - lr: 1.0000e-04\n",
            "Epoch 5/50\n",
            "75/75 [==============================] - 71s 943ms/step - loss: 0.1664 - peak_signal_noise_ratio: 63.5782 - val_loss: 0.1216 - val_peak_signal_noise_ratio: 65.8809 - lr: 1.0000e-04\n",
            "Epoch 6/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1749 - peak_signal_noise_ratio: 63.4952 - val_loss: 0.1197 - val_peak_signal_noise_ratio: 66.0266 - lr: 1.0000e-04\n",
            "Epoch 7/50\n",
            "75/75 [==============================] - 71s 942ms/step - loss: 0.1644 - peak_signal_noise_ratio: 63.7675 - val_loss: 0.1204 - val_peak_signal_noise_ratio: 65.8658 - lr: 1.0000e-04\n",
            "Epoch 8/50\n",
            "75/75 [==============================] - 71s 943ms/step - loss: 0.1542 - peak_signal_noise_ratio: 64.2077 - val_loss: 0.1153 - val_peak_signal_noise_ratio: 66.4227 - lr: 1.0000e-04\n",
            "Epoch 9/50\n",
            "75/75 [==============================] - 80s 1s/step - loss: 0.1598 - peak_signal_noise_ratio: 63.9319 - val_loss: 0.1176 - val_peak_signal_noise_ratio: 66.4509 - lr: 1.0000e-04\n",
            "Epoch 10/50\n",
            "75/75 [==============================] - 80s 1s/step - loss: 0.1576 - peak_signal_noise_ratio: 64.0152 - val_loss: 0.1159 - val_peak_signal_noise_ratio: 66.3044 - lr: 1.0000e-04\n",
            "Epoch 11/50\n",
            "75/75 [==============================] - 70s 941ms/step - loss: 0.1525 - peak_signal_noise_ratio: 64.3085 - val_loss: 0.1184 - val_peak_signal_noise_ratio: 66.5723 - lr: 1.0000e-04\n",
            "Epoch 12/50\n",
            "75/75 [==============================] - 70s 941ms/step - loss: 0.1465 - peak_signal_noise_ratio: 64.6508 - val_loss: 0.1136 - val_peak_signal_noise_ratio: 66.8768 - lr: 1.0000e-04\n",
            "Epoch 13/50\n",
            "75/75 [==============================] - 80s 1s/step - loss: 0.1455 - peak_signal_noise_ratio: 64.8678 - val_loss: 0.1234 - val_peak_signal_noise_ratio: 66.4972 - lr: 1.0000e-04\n",
            "Epoch 14/50\n",
            "75/75 [==============================] - 71s 942ms/step - loss: 0.1473 - peak_signal_noise_ratio: 64.8383 - val_loss: 0.1146 - val_peak_signal_noise_ratio: 66.8325 - lr: 1.0000e-04\n",
            "Epoch 15/50\n",
            "75/75 [==============================] - 71s 942ms/step - loss: 0.1467 - peak_signal_noise_ratio: 64.9732 - val_loss: 0.1084 - val_peak_signal_noise_ratio: 67.1350 - lr: 1.0000e-04\n",
            "Epoch 16/50\n",
            "75/75 [==============================] - 70s 937ms/step - loss: 0.1482 - peak_signal_noise_ratio: 64.7597 - val_loss: 0.1158 - val_peak_signal_noise_ratio: 66.9836 - lr: 1.0000e-04\n",
            "Epoch 17/50\n",
            "75/75 [==============================] - 79s 1s/step - loss: 0.1439 - peak_signal_noise_ratio: 64.9124 - val_loss: 0.1150 - val_peak_signal_noise_ratio: 66.9349 - lr: 1.0000e-04\n",
            "Epoch 18/50\n",
            "75/75 [==============================] - 71s 942ms/step - loss: 0.1402 - peak_signal_noise_ratio: 65.1801 - val_loss: 0.1096 - val_peak_signal_noise_ratio: 67.3645 - lr: 1.0000e-04\n",
            "Epoch 19/50\n",
            "75/75 [==============================] - 70s 937ms/step - loss: 0.1395 - peak_signal_noise_ratio: 65.2925 - val_loss: 0.1113 - val_peak_signal_noise_ratio: 67.1285 - lr: 1.0000e-04\n",
            "Epoch 20/50\n",
            "75/75 [==============================] - 80s 1s/step - loss: 0.1369 - peak_signal_noise_ratio: 65.4434 - val_loss: 0.1256 - val_peak_signal_noise_ratio: 66.5258 - lr: 1.0000e-04\n",
            "Epoch 21/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1400 - peak_signal_noise_ratio: 65.2900 - val_loss: 0.1166 - val_peak_signal_noise_ratio: 67.0571 - lr: 1.0000e-04\n",
            "Epoch 22/50\n",
            "75/75 [==============================] - 71s 943ms/step - loss: 0.1435 - peak_signal_noise_ratio: 65.2407 - val_loss: 0.1231 - val_peak_signal_noise_ratio: 66.5538 - lr: 1.0000e-04\n",
            "Epoch 23/50\n",
            "75/75 [==============================] - ETA: 0s - loss: 0.1414 - peak_signal_noise_ratio: 65.1852\n",
            "Epoch 23: ReduceLROnPlateau reducing learning rate to 4.999999873689376e-05.\n",
            "75/75 [==============================] - 71s 940ms/step - loss: 0.1414 - peak_signal_noise_ratio: 65.1852 - val_loss: 0.1236 - val_peak_signal_noise_ratio: 66.7622 - lr: 1.0000e-04\n",
            "Epoch 24/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1454 - peak_signal_noise_ratio: 64.9235 - val_loss: 0.1069 - val_peak_signal_noise_ratio: 67.4382 - lr: 5.0000e-05\n",
            "Epoch 25/50\n",
            "75/75 [==============================] - 70s 937ms/step - loss: 0.1361 - peak_signal_noise_ratio: 65.5833 - val_loss: 0.1065 - val_peak_signal_noise_ratio: 67.5202 - lr: 5.0000e-05\n",
            "Epoch 26/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1337 - peak_signal_noise_ratio: 65.7435 - val_loss: 0.1096 - val_peak_signal_noise_ratio: 67.2198 - lr: 5.0000e-05\n",
            "Epoch 27/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1305 - peak_signal_noise_ratio: 65.8251 - val_loss: 0.1027 - val_peak_signal_noise_ratio: 67.6947 - lr: 5.0000e-05\n",
            "Epoch 28/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1242 - peak_signal_noise_ratio: 66.1865 - val_loss: 0.1139 - val_peak_signal_noise_ratio: 66.7709 - lr: 5.0000e-05\n",
            "Epoch 29/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1225 - peak_signal_noise_ratio: 66.4003 - val_loss: 0.1127 - val_peak_signal_noise_ratio: 66.7394 - lr: 5.0000e-05\n",
            "Epoch 30/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1236 - peak_signal_noise_ratio: 66.2518 - val_loss: 0.1041 - val_peak_signal_noise_ratio: 67.4816 - lr: 5.0000e-05\n",
            "Epoch 31/50\n",
            "75/75 [==============================] - 70s 941ms/step - loss: 0.1278 - peak_signal_noise_ratio: 66.0467 - val_loss: 0.1036 - val_peak_signal_noise_ratio: 67.5849 - lr: 5.0000e-05\n",
            "Epoch 32/50\n",
            "75/75 [==============================] - ETA: 0s - loss: 0.1285 - peak_signal_noise_ratio: 65.9661\n",
            "Epoch 32: ReduceLROnPlateau reducing learning rate to 2.499999936844688e-05.\n",
            "75/75 [==============================] - 70s 938ms/step - loss: 0.1285 - peak_signal_noise_ratio: 65.9661 - val_loss: 0.1032 - val_peak_signal_noise_ratio: 67.5509 - lr: 5.0000e-05\n",
            "Epoch 33/50\n",
            "75/75 [==============================] - 70s 934ms/step - loss: 0.1193 - peak_signal_noise_ratio: 66.6949 - val_loss: 0.1090 - val_peak_signal_noise_ratio: 67.3078 - lr: 2.5000e-05\n",
            "Epoch 34/50\n",
            "75/75 [==============================] - 70s 941ms/step - loss: 0.1191 - peak_signal_noise_ratio: 66.6520 - val_loss: 0.1056 - val_peak_signal_noise_ratio: 67.3013 - lr: 2.5000e-05\n",
            "Epoch 35/50\n",
            "75/75 [==============================] - 70s 941ms/step - loss: 0.1171 - peak_signal_noise_ratio: 66.8241 - val_loss: 0.1097 - val_peak_signal_noise_ratio: 66.9988 - lr: 2.5000e-05\n",
            "Epoch 36/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1191 - peak_signal_noise_ratio: 66.5754 - val_loss: 0.1153 - val_peak_signal_noise_ratio: 66.6156 - lr: 2.5000e-05\n",
            "Epoch 37/50\n",
            "75/75 [==============================] - ETA: 0s - loss: 0.1199 - peak_signal_noise_ratio: 66.5940\n",
            "Epoch 37: ReduceLROnPlateau reducing learning rate to 1.249999968422344e-05.\n",
            "75/75 [==============================] - 70s 932ms/step - loss: 0.1199 - peak_signal_noise_ratio: 66.5940 - val_loss: 0.1120 - val_peak_signal_noise_ratio: 66.7965 - lr: 2.5000e-05\n",
            "Epoch 38/50\n",
            "75/75 [==============================] - 70s 936ms/step - loss: 0.1189 - peak_signal_noise_ratio: 66.6848 - val_loss: 0.1063 - val_peak_signal_noise_ratio: 67.5486 - lr: 1.2500e-05\n",
            "Epoch 39/50\n",
            "75/75 [==============================] - 70s 941ms/step - loss: 0.1156 - peak_signal_noise_ratio: 66.8729 - val_loss: 0.1073 - val_peak_signal_noise_ratio: 67.2537 - lr: 1.2500e-05\n",
            "Epoch 40/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1150 - peak_signal_noise_ratio: 66.8736 - val_loss: 0.1066 - val_peak_signal_noise_ratio: 67.3235 - lr: 1.2500e-05\n",
            "Epoch 41/50\n",
            "75/75 [==============================] - 70s 935ms/step - loss: 0.1135 - peak_signal_noise_ratio: 66.9386 - val_loss: 0.1103 - val_peak_signal_noise_ratio: 67.0231 - lr: 1.2500e-05\n",
            "Epoch 42/50\n",
            "75/75 [==============================] - 71s 941ms/step - loss: 0.1130 - peak_signal_noise_ratio: 67.0489 - val_loss: 0.1001 - val_peak_signal_noise_ratio: 67.7598 - lr: 1.2500e-05\n",
            "Epoch 43/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1133 - peak_signal_noise_ratio: 67.0871 - val_loss: 0.1120 - val_peak_signal_noise_ratio: 66.7947 - lr: 1.2500e-05\n",
            "Epoch 44/50\n",
            "75/75 [==============================] - 70s 933ms/step - loss: 0.1134 - peak_signal_noise_ratio: 67.1683 - val_loss: 0.1106 - val_peak_signal_noise_ratio: 67.0630 - lr: 1.2500e-05\n",
            "Epoch 45/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1168 - peak_signal_noise_ratio: 66.8179 - val_loss: 0.1091 - val_peak_signal_noise_ratio: 67.1748 - lr: 1.2500e-05\n",
            "Epoch 46/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1073 - peak_signal_noise_ratio: 67.5175 - val_loss: 0.1007 - val_peak_signal_noise_ratio: 67.7764 - lr: 1.2500e-05\n",
            "Epoch 47/50\n",
            "75/75 [==============================] - 70s 937ms/step - loss: 0.1149 - peak_signal_noise_ratio: 66.9879 - val_loss: 0.1073 - val_peak_signal_noise_ratio: 67.5095 - lr: 1.2500e-05\n",
            "Epoch 48/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1124 - peak_signal_noise_ratio: 67.1661 - val_loss: 0.1074 - val_peak_signal_noise_ratio: 67.4590 - lr: 1.2500e-05\n",
            "Epoch 49/50\n",
            "75/75 [==============================] - 70s 939ms/step - loss: 0.1113 - peak_signal_noise_ratio: 67.2660 - val_loss: 0.1060 - val_peak_signal_noise_ratio: 67.4733 - lr: 1.2500e-05\n",
            "Epoch 50/50\n",
            "75/75 [==============================] - 70s 940ms/step - loss: 0.1117 - peak_signal_noise_ratio: 67.2173 - val_loss: 0.1090 - val_peak_signal_noise_ratio: 67.3992 - lr: 1.2500e-05\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"
          }
        },
        {
          "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": [
        "def plot_results(images, titles, figure_size=(12, 12)):\n",
        "    fig = plt.figure(figsize=figure_size)\n",
        "    for i in range(len(images)):\n",
        "        fig.add_subplot(1, len(images), i + 1).set_title(titles[i])\n",
        "        _ = plt.imshow(images[i])\n",
        "        plt.axis(\"off\")\n",
        "    plt.show()\n",
        "\n",
        "\n",
        "def infer(original_image):\n",
        "    image = keras.preprocessing.image.img_to_array(original_image)\n",
        "    image = image.astype(\"float32\") / 255.0\n",
        "    image = np.expand_dims(image, axis=0)\n",
        "    output = model.predict(image)\n",
        "    output_image = output[0] * 255.0\n",
        "    output_image = output_image.clip(0, 255)\n",
        "    output_image = output_image.reshape(\n",
        "        (np.shape(output_image)[0], np.shape(output_image)[1], 3)\n",
        "    )\n",
        "    output_image = Image.fromarray(np.uint8(output_image))\n",
        "    original_image = Image.fromarray(np.uint8(original_image))\n",
        "    return output_image"
      ],
      "metadata": {
        "id": "Xil5G_Sde9FR"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.save('MIRNet_v2.hdf5')"
      ],
      "metadata": {
        "id": "rdVVVNkz72v9"
      },
      "execution_count": 12,
      "outputs": []
    }
  ]
}