training_part_1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "machine_shape": "hm",
      "gpuType": "L4"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "!wget https://isic-challenge-data.s3.amazonaws.com/2016/ISBI2016_ISIC_Part1_Training_Data.zip -O training.zip\n",
        "!wget https://isic-challenge-data.s3.amazonaws.com/2016/ISBI2016_ISIC_Part1_Training_GroundTruth.zip -O training_truth.zip\n",
        "!wget https://isic-challenge-data.s3.amazonaws.com/2016/ISBI2016_ISIC_Part1_Test_Data.zip -O testing.zip\n",
        "!wget https://isic-challenge-data.s3.amazonaws.com/2016/ISBI2016_ISIC_Part1_Test_GroundTruth.zip -O testing_truth.zip"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gmJ3n4n6sO_n",
        "outputId": "a988ea88-65c7-459e-c2e7-829a46f3a5a0"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2024-11-13 23:21:13--  https://isic-challenge-data.s3.amazonaws.com/2016/ISBI2016_ISIC_Part1_Test_Data.zip\n",
            "Resolving isic-challenge-data.s3.amazonaws.com (isic-challenge-data.s3.amazonaws.com)... 54.231.225.201, 54.231.130.177, 52.217.232.97, ...\n",
            "Connecting to isic-challenge-data.s3.amazonaws.com (isic-challenge-data.s3.amazonaws.com)|54.231.225.201|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 242932767 (232M) [application/zip]\n",
            "Saving to: ‘testing.zip’\n",
            "\n",
            "testing.zip         100%[===================>] 231.68M  13.9MB/s    in 18s     \n",
            "\n",
            "2024-11-13 23:21:33 (12.6 MB/s) - ‘testing.zip’ saved [242932767/242932767]\n",
            "\n",
            "--2024-11-13 23:21:33--  https://isic-challenge-data.s3.amazonaws.com/2016/ISBI2016_ISIC_Part1_Test_GroundTruth.zip\n",
            "Resolving isic-challenge-data.s3.amazonaws.com (isic-challenge-data.s3.amazonaws.com)... 3.5.29.43, 52.217.67.60, 3.5.29.209, ...\n",
            "Connecting to isic-challenge-data.s3.amazonaws.com (isic-challenge-data.s3.amazonaws.com)|3.5.29.43|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 2376407 (2.3M) [application/zip]\n",
            "Saving to: ‘testing_truth.zip’\n",
            "\n",
            "testing_truth.zip   100%[===================>]   2.27M  1.49MB/s    in 1.5s    \n",
            "\n",
            "2024-11-13 23:21:36 (1.49 MB/s) - ‘testing_truth.zip’ saved [2376407/2376407]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!unzip -qq -n training.zip\n",
        "!unzip -qq -n training_truth.zip\n",
        "!unzip -qq -n testing.zip\n",
        "!unzip -qq -n testing_truth.zip"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DKpD6y9DtmOh",
        "outputId": "00a9e2d0-2463-4be3-8955-ce0d6dfbfe74"
      },
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "unzip:  cannot find or open training_truth.zip, training_truth.zip.zip or training_truth.zip.ZIP.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import random\n",
        "import gc\n",
        "\n",
        "import numpy\n",
        "import tensorflow as tf\n",
        "from tensorflow import data\n",
        "from keras import layers, utils, losses, metrics\n",
        "from matplotlib import pyplot\n",
        "import seaborn\n",
        "import pandas\n",
        "from tqdm import tqdm\n",
        "\n",
        "# Check for GPU\n",
        "print(tf.config.list_physical_devices(\"GPU\"))\n",
        "\n",
        "# For seed sync\n",
        "random.seed()\n",
        "\n",
        "RAW_WIDTH = 512\n",
        "RAW_HEIGHT = 512\n",
        "RAW_SIZE = (RAW_WIDTH, RAW_HEIGHT)\n",
        "WIDTH = 256\n",
        "HEIGHT = 256\n",
        "SIZE = (WIDTH, HEIGHT)\n",
        "TRAINING_DIR = \"/content/ISBI2016_ISIC_Part1_Training_Data\"\n",
        "TRAINING_TRUTH_DIR = \"/content/ISBI2016_ISIC_Part1_Training_GroundTruth\"\n",
        "BATCH_SIZE = 200\n",
        "SAVED_FILE = f\"p1_weights_{WIDTH}_{HEIGHT}.h5\""
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_mEN2id6tK1B",
        "outputId": "4b439d26-bf75-431c-d8dc-45f934e81aba"
      },
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "seed = random.randint(0, 2**32 - 1)\n",
        "\n",
        "input_images = utils.image_dataset_from_directory(\n",
        "    TRAINING_DIR,\n",
        "    labels=None,\n",
        "    label_mode=None,\n",
        "    image_size=RAW_SIZE,\n",
        "    batch_size=BATCH_SIZE,\n",
        "    color_mode=\"rgb\",\n",
        "    interpolation=\"bilinear\",\n",
        "    seed=seed\n",
        ")\n",
        "\n",
        "input_images = input_images.cache()\n",
        "\n",
        "label_images = utils.image_dataset_from_directory(\n",
        "    TRAINING_TRUTH_DIR,\n",
        "    labels=None,\n",
        "    label_mode=None,\n",
        "    image_size=RAW_SIZE,\n",
        "    batch_size=BATCH_SIZE,\n",
        "    color_mode=\"grayscale\",\n",
        "    interpolation=\"nearest\",\n",
        "    seed=seed\n",
        ")\n",
        "\n",
        "label_images = label_images.cache()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TxAp7JgZtik9",
        "outputId": "d12522d3-4b76-4782-ddbf-13b65e60536b"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 900 files.\n",
            "Found 900 files.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class Preprocessing():\n",
        "    def __init__(self):\n",
        "        seed = random.randint(0, 2**32 - 1)\n",
        "        self.scale = layers.Rescaling(scale=1./255)\n",
        "        self.rotate_x = layers.RandomRotation(\n",
        "            factor=0.2, fill_mode=\"constant\", interpolation=\"bilinear\", seed=seed)\n",
        "        self.rotate_y = layers.RandomRotation(\n",
        "            factor=0.2, fill_mode=\"constant\", interpolation=\"nearest\", seed=seed)\n",
        "        self.crop_x = layers.RandomCrop(height=HEIGHT, width=WIDTH, seed=seed)\n",
        "        self.crop_y = layers.RandomCrop(height=HEIGHT, width=WIDTH, seed=seed)\n",
        "        self.brightness = layers.RandomBrightness(\n",
        "            factor=0.2, value_range=(0, 1))\n",
        "        self.contrast = layers.RandomContrast(factor=0.2)\n",
        "        self.flip_x = layers.RandomFlip(seed=seed)\n",
        "        self.flip_y = layers.RandomFlip(seed=seed)\n",
        "\n",
        "    def __call__(self, inputs, labels):\n",
        "        index = numpy.arange(inputs.shape[0] * 2)\n",
        "        numpy.random.shuffle(index)\n",
        "        # Inputs\n",
        "        x = self.scale(inputs)\n",
        "        x = self.rotate_x(x)\n",
        "        x = self.crop_x(x)\n",
        "        x = tf.repeat(x, repeats=2, axis=0)\n",
        "        x = self.brightness(x)\n",
        "        x = self.contrast(x)\n",
        "        x = self.flip_x(x)\n",
        "        x = tf.convert_to_tensor(x.numpy()[index])\n",
        "        # Labels\n",
        "        y = self.scale(labels)\n",
        "        y = self.rotate_y(y)\n",
        "        y = self.crop_y(y)\n",
        "        y = tf.repeat(y, repeats=2, axis=0)\n",
        "        y = self.flip_y(y)\n",
        "        y = tf.convert_to_tensor(y.numpy()[index], dtype=\"int8\")\n",
        "        return (x, y)\n",
        "\n",
        "\n",
        "preprocessing = Preprocessing()"
      ],
      "metadata": {
        "id": "qJt_Bc48wFYv"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from model import UNet\n",
        "\n",
        "model = UNet()\n",
        "model.compile(\n",
        "    optimizer=tf.keras.optimizers.Adam(learning_rate=1e-4),\n",
        "    loss=losses.BinaryFocalCrossentropy(\n",
        "        apply_class_balancing=False,\n",
        "        alpha=0.25,\n",
        "        gamma=2.0\n",
        "    ),\n",
        "    metrics=[\n",
        "        \"BinaryAccuracy\",\n",
        "        \"BinaryIoU\"\n",
        "    ]\n",
        ")\n",
        "model(layers.Input((WIDTH, HEIGHT, 3)))\n",
        "iou_history = []"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ialA2Kgyv5H3",
        "outputId": "e8a85520-1a0c-4f98-f21e-e4a65619ea00"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<KerasTensor shape=(None, 256, 256, 1), dtype=float32, sparse=False, name=keras_tensor_37>"
            ]
          },
          "metadata": {},
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "EPOCH = 25\n",
        "\n",
        "for epoch in range(EPOCH):\n",
        "    print(f\"=== Epoch {epoch} ===\")\n",
        "    zipped = tf.data.Dataset.zip((input_images, label_images))\n",
        "    for (inputs, labels) in tqdm(zipped):\n",
        "        inputs, labels = preprocessing(inputs, labels)\n",
        "        history = model.fit(inputs, labels, verbose=0)\n",
        "        iou_history.append(history.history[\"BinaryIoU\"])\n",
        "    gc.collect()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "collapsed": true,
        "id": "rzOngTSswKL5",
        "outputId": "d3cc678b-0c4d-4b27-d706-2c76bdd3c227"
      },
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 0 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 1 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 2 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 3 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 4 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 5 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 6 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 7 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 8 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.01it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 9 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 10 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 11 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 12 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 13 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 14 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 15 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 16 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 17 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 18 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 19 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:08<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 20 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 21 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 22 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 23 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  2.00it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "=== Epoch 24 ===\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 18/18 [00:09<00:00,  1.99it/s]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "iou_history_pd = pandas.DataFrame(iou_history)\n",
        "seaborn.lineplot(iou_history_pd)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 453
        },
        "id": "oXy_506-x1hx",
        "outputId": "5f47fe07-bb0d-46a0-aec2-5fea51543110"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 26
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "TRAINING_DIR = \"/content/ISBI2016_ISIC_Part1_Training_Data\"\n",
        "TRAINING_TRUTH_DIR = \"/content/ISBI2016_ISIC_Part1_Training_GroundTruth\"\n",
        "BATCH_SIZE = 1\n",
        "SAVED_FILE = f\"saved/v1/weights_{WIDTH}_{HEIGHT}.h5\""
      ],
      "metadata": {
        "id": "4WpKo5xu1mwf"
      },
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "seed = 100\n",
        "\n",
        "input_images = utils.image_dataset_from_directory(\n",
        "    TRAINING_DIR,\n",
        "    labels=None,\n",
        "    label_mode=None,\n",
        "    image_size=SIZE,\n",
        "    batch_size=BATCH_SIZE,\n",
        "    color_mode=\"rgb\",\n",
        "    shuffle=False,\n",
        "    seed=seed\n",
        ")\n",
        "\n",
        "input_images = input_images.cache()\n",
        "\n",
        "label_images = utils.image_dataset_from_directory(\n",
        "    TRAINING_TRUTH_DIR,\n",
        "    labels=None,\n",
        "    label_mode=None,\n",
        "    image_size=SIZE,\n",
        "    batch_size=BATCH_SIZE,\n",
        "    color_mode=\"grayscale\",\n",
        "    shuffle=False,\n",
        "    seed=seed\n",
        ")\n",
        "\n",
        "label_images = label_images.cache()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hFi-ycbP1qMA",
        "outputId": "eb140d67-90e2-49e0-83eb-68e7ff9df1b2"
      },
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 900 files.\n",
            "Found 900 files.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "metric = metrics.BinaryIoU()\n",
        "loss_list = []\n",
        "for (inputs, labels) in tqdm(data.Dataset.zip((input_images, label_images))):\n",
        "    labels = tf.cast(labels > 0.5, dtype=\"int8\")\n",
        "    outputs = model(inputs / 255)\n",
        "    loss_value = metric(labels, outputs).numpy()\n",
        "    loss_list.append(loss_value)\n",
        "total_loss = numpy.average(loss_list)\n",
        "print(f\"Mean Itersect-over-Union: {total_loss}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_tLXLZgY1vKu",
        "outputId": "a2fc85b1-7697-4cfe-db57-6f26a205910b"
      },
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 900/900 [00:53<00:00, 16.76it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mean Itersect-over-Union: 0.7138893008232117\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "jQRHwX3k2eU2"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}