tumor_classification.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "# Install Libraries\n",
        "!pip install -q kaggle\n",
        "!pip install torchmetrics\n",
        "\n",
        "# Import Libraries\n",
        "from google.colab import drive, files\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torchvision\n",
        "from torch.utils.data import Subset\n",
        "import torch.optim as optim\n",
        "from torch.optim import lr_scheduler\n",
        "from torchvision import datasets, models, transforms\n",
        "from torch.utils.data.sampler import SubsetRandomSampler\n",
        "import numpy as np\n",
        "from torchmetrics import ConfusionMatrix, ROC, AUROC\n",
        "from torchmetrics.functional import auc\n",
        "from sklearn.metrics import classification_report, roc_curve, auc\n",
        "import seaborn as sns\n",
        "import time\n",
        "import os\n",
        "import copy\n",
        "import matplotlib.pyplot as plt\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jbCu2X2noI7A",
        "outputId": "595380e5-7f14-4394-b676-85a28ad59e45"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting torchmetrics\n",
            "  Downloading torchmetrics-0.9.3-py3-none-any.whl (419 kB)\n",
            "\u001b[K     |████████████████████████████████| 419 kB 4.1 MB/s \n",
            "\u001b[?25hRequirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torchmetrics) (4.1.1)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.7/dist-packages (from torchmetrics) (21.3)\n",
            "Requirement already satisfied: numpy>=1.17.2 in /usr/local/lib/python3.7/dist-packages (from torchmetrics) (1.21.6)\n",
            "Requirement already satisfied: torch>=1.3.1 in /usr/local/lib/python3.7/dist-packages (from torchmetrics) (1.12.1+cu113)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging->torchmetrics) (3.0.9)\n",
            "Installing collected packages: torchmetrics\n",
            "Successfully installed torchmetrics-0.9.3\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1-UwOy-Kn5yV",
        "outputId": "3521b636-c2de-4f0d-a63f-c3c0ebddc3c4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n",
            "[INFO]: Google Drive connected.\n",
            "/content/drive/MyDrive/colab/tumor_classification\n"
          ]
        }
      ],
      "source": [
        "# Connect to Google Drive\n",
        "drive.mount('/content/drive')\n",
        "print(\"[INFO]: Google Drive connected.\")\n",
        "%cd ./drive/MyDrive/colab/tumor_classification/"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "HOW_86Ppn_-H"
      },
      "outputs": [],
      "source": [
        "# Download the dataset\n",
        "#files.upload()\n",
        "!mkdir ~/.kaggle\n",
        "!cp kaggle.json ~/.kaggle\n",
        "!chmod 600 ~/.kaggle/kaggle.json\n",
        "!kaggle datasets download -d masoudnickparvar/brain-tumor-mri-dataset #sartajbhuvaji/brain-tumor-classification-mri\n",
        "!mkdir data2\n",
        "!unzip brain-tumor-mri-dataset -d data2"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yeAk-9N3n_n-",
        "outputId": "87bb2288-236c-4eb7-c406-bc65ce88cbc4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Testing  Training\n"
          ]
        }
      ],
      "source": [
        "# Check for dataset\n",
        "!ls data2"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Combine Testing and Training Data\n",
        "!rsync -a data2/Testing/ data2/Training/"
      ],
      "metadata": {
        "id": "V17jbCAyEHtH"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gOW-Q09VAu62",
        "outputId": "2b75cdd1-1968-40bc-ecf3-f8e636ca3cf7"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[INFO]: Running on the GPU\n",
            "[GPU INFO]: _CudaDeviceProperties(name='Tesla T4', major=7, minor=5, total_memory=15109MB, multi_processor_count=40)\n"
          ]
        }
      ],
      "source": [
        "# Look for available GPU\n",
        "if torch.cuda.is_available():\n",
        "    device = torch.device(\"cuda:0\")\n",
        "    print(\"[INFO]: Running on the GPU\")\n",
        "    print(f\"[GPU INFO]: %s\" % torch.cuda.get_device_properties(device))\n",
        "else:\n",
        "    device = torch.device(\"cpu\")\n",
        "    print(\"[INFO]: Running on the CPU\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vujUH4FChoWe",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c348f5ba-b9ad-440e-958f-7dcf3ec6106d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[INFO]: Total Size: 3160  Train Size: 2844  Test Size: 316\n",
            "[INFO]: Number of classes: 4  Class names: ['glioma_tumor', 'meningioma_tumor', 'no_tumor', 'pituitary_tumor']\n"
          ]
        }
      ],
      "source": [
        "# Prepare the data\n",
        "\n",
        "random_seed= 42\n",
        "validation_split = .1\n",
        "shuffle_dataset = True\n",
        "image_size = 128\n",
        "batch_size = 32\n",
        "num_workers = 2\n",
        "\n",
        "preprocess = transforms.Compose([\n",
        "                transforms.Resize(image_size),\n",
        "                transforms.CenterCrop((image_size,image_size)),\n",
        "                transforms.ToTensor(),\n",
        "                #transforms.Grayscale(),\n",
        "                transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),\n",
        "                #transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),\n",
        "                transforms.RandomHorizontalFlip(),\n",
        "            ])\n",
        "trainset=torchvision.datasets.ImageFolder(root='data/Training/', transform=preprocess)\n",
        "#testset=torchvision.datasets.ImageFolder(root='data/Testing/', transform=preprocess)\n",
        "\n",
        "dataset_size = len(trainset.imgs)\n",
        "test_size = int(0.1 * dataset_size)\n",
        "train_size = int(dataset_size - test_size)\n",
        "print(f'[INFO]: Total Size: {dataset_size}  Train Size: {train_size}  Test Size: {test_size}')\n",
        "\n",
        "indices = list(range(dataset_size))\n",
        "split = int(np.floor(validation_split * dataset_size))\n",
        "if shuffle_dataset :\n",
        "    np.random.seed(random_seed)\n",
        "    np.random.shuffle(indices)\n",
        "train_indices, val_indices = indices[split:], indices[:split]\n",
        "\n",
        "train_sampler = SubsetRandomSampler(train_indices)\n",
        "valid_sampler = SubsetRandomSampler(val_indices)\n",
        "\n",
        "trainloader = torch.utils.data.DataLoader(trainset, batch_size=32, num_workers=2, sampler=train_sampler)\n",
        "validationloader = torch.utils.data.DataLoader(trainset, batch_size=train_size, num_workers=2, sampler=valid_sampler)\n",
        "\n",
        "\n",
        "class_names = trainset.classes\n",
        "num_classes = len(class_names)\n",
        "print(f'[INFO]: Number of classes: {num_classes}  Class names: {class_names}')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "lLL-Uzp0CX1f"
      },
      "outputs": [],
      "source": [
        "%matplotlib inline\n",
        "\n",
        "def save_model(model):\n",
        "    # Save the model\n",
        "    path = \"./tc_\"+ model.name +\".pt\"\n",
        "    torch.save(model, path)\n",
        "\n",
        "\n",
        "def test_model(model,loader):\n",
        "    # Test the model\n",
        "    correct = 0\n",
        "    total = 0\n",
        "    with torch.no_grad():\n",
        "      for data in loader:\n",
        "        images, labels = data\n",
        "        images = images.to(device)\n",
        "        labels = labels.to(device)\n",
        "\n",
        "        outputs = model(images)\n",
        "\n",
        "        pred_softmax = torch.log_softmax(outputs.data, dim = 1)\n",
        "        _, predicted = torch.max(pred_softmax, dim = 1)\n",
        "\n",
        "        total += labels.size(0)\n",
        "        correct += (predicted == labels).sum().item()\n",
        "\n",
        "        probabilities = predicted\n",
        "        predictions = predicted.cpu()\n",
        "        all_labels = labels.cpu()\n",
        "\n",
        "      c_report = classification_report(all_labels, predictions, target_names=class_names)\n",
        "      print(c_report)\n",
        "\n",
        "      confmat = ConfusionMatrix(num_classes=num_classes)\n",
        "      heatmap = confmat(predictions, all_labels)\n",
        "      sns.set_theme()\n",
        "      heatmap = heatmap.cpu().detach().numpy()\n",
        "      ax = sns.heatmap(heatmap, annot=True, fmt=\"d\", yticklabels=class_names, xticklabels=class_names, cmap=\"YlGnBu\")\n",
        "      fig = ax.get_figure()\n",
        "      fig.tight_layout()\n",
        "      heatmap_name = 'heatmap_' + model.name + '.png'\n",
        "      print(heatmap_name)\n",
        "      fig.savefig(heatmap_name,dpi=300) \n",
        "      \n",
        "    accuracy = round(correct/total, 5)\n",
        "    print(f'Accuracy: {accuracy}')\n",
        "\n",
        "    with open('tc_statistics.txt', 'w') as f:\n",
        "      f.write(c_report + '\\n\\nAccuracy:' + str(accuracy))\n",
        "\n",
        "    return accuracy\n",
        "\n",
        "valid_losses = []\n",
        "train_losses = []\n",
        "\n",
        "def train_model(model, criterion, optimizer, scheduler, num_epochs=25):\n",
        "    # Train the model\n",
        "    since = time.time()\n",
        "\n",
        "    best_accuracy = 0.0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print(f'Epoch {epoch+1}/{num_epochs}')\n",
        "        print('-' * 20)\n",
        "\n",
        "        ## Training\n",
        "        model.train()\n",
        "        train_loss = 0\n",
        "        for data, labels in trainloader:\n",
        "            data = data.to(device)\n",
        "            labels = labels.to(device)\n",
        "            optimizer.zero_grad()\n",
        "            output = model(data)\n",
        "            loss = criterion(output, labels)\n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "            train_loss += loss.item()\n",
        "        ## Validation\n",
        "        model.eval()\n",
        "        valid_loss = 0\n",
        "        correct = 0\n",
        "        total = 0\n",
        "\n",
        "        with torch.no_grad():\n",
        "            for data in validationloader:\n",
        "              images, labels = data\n",
        "              images = images.to(device)\n",
        "              labels = labels.to(device)\n",
        "\n",
        "              outputs = model(images)\n",
        "\n",
        "              pred_softmax = torch.log_softmax(outputs.data, dim = 1)\n",
        "              _, predicted = torch.max(pred_softmax, dim = 1)\n",
        "\n",
        "              total += labels.size(0)\n",
        "              correct += (predicted == labels).sum().item()\n",
        "              loss = criterion(outputs, labels) \n",
        "              valid_loss += loss.item()\n",
        "              \n",
        "\n",
        "              accuracy = round(correct/total, 5)\n",
        "\n",
        "              if accuracy > best_accuracy:\n",
        "                save_model(model)\n",
        "                best_accuracy = accuracy\n",
        "                print(f'[INFO]: Model saved. Accuracy: {accuracy}')\n",
        "                \n",
        "        \n",
        "        train_loss = train_loss / len(trainloader)\n",
        "        valid_loss = valid_loss / len(validationloader)\n",
        "        train_losses.append(train_loss)\n",
        "        valid_losses.append(valid_loss)\n",
        "        print(f'[EPOCH]: Training loss: {train_loss}   Validation Loss: {valid_loss}')\n",
        "        print('-' * 90)\n",
        "        plot_loss(train_losses,valid_losses,phase='epoch')\n",
        "        \n",
        "    time_elapsed = time.time() - since\n",
        "    print(f'[TRAINING]: Training complete in {time_elapsed // 60:.0f}m {time_elapsed % 60:.0f}s')\n",
        "    print(f'[TRAINING]: Best val Accuracy: {best_accuracy:4f}')\n",
        "\n",
        "    plot_loss(train_losses,valid_losses,phase='final')\n",
        "\n",
        "def plot_loss(trainloss,validationloss,phase):\n",
        "  plt.figure(figsize=(10,5))\n",
        "  plt.title(\"Training and Validation Loss\")\n",
        "  plt.plot(validationloss,label=\"val\")\n",
        "  plt.plot(trainloss,label=\"train\")\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(\"Loss\")\n",
        "  plt.legend()\n",
        "  if phase == 'final':\n",
        "    plt.savefig('loss_graph.png', dpi=300)\n",
        "    plt.show()\n",
        "  elif phase == 'epoch':\n",
        "    plt.savefig('loss_graph.png', dpi=300)\n",
        "    plt.close('all')\n",
        "  "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "k0zRgeYymnH1",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fcd41341-d008-488e-8e53-afbc0d70c19b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[INFO]: Loading pre-trained weights...\n",
            "[INFO]: Fine-tuning all layers...\n"
          ]
        }
      ],
      "source": [
        "# Import pretrained models\n",
        "\n",
        "def build_model(model_name, weights, pretrained=True, fine_tune=True, num_classes=2):\n",
        "\n",
        "    # Models which have fc instead of classifier\n",
        "    fc = ['googlenet',\n",
        "          'regnet_y_400mf',\n",
        "          'regnet_y_800mf',\n",
        "          'regnet_y_1_6gf',\n",
        "          'regnet_y_3_2gf',\n",
        "          'regnet_y_8gf',\n",
        "          'regnet_y_16gf',\n",
        "          'regnet_y_32gf',\n",
        "          'regnet_y_128gf',\n",
        "          'regnet_x_400mf',\n",
        "          'regnet_x_800mf',\n",
        "          'regnet_x_1_6gf',\n",
        "          'regnet_x_3_2gf',\n",
        "          'regnet_x_8gf',\n",
        "          'regnet_x_16gf',\n",
        "          'regnet_x_32gf',\n",
        "          'resnet18',\n",
        "          'resnet34',\n",
        "          'resnet50',\n",
        "          'resnet101',\n",
        "          'resnet152',\n",
        "          'resnext50_32x4d',\n",
        "          'resnext101_32x8d',\n",
        "          'resnext101_64x4d',\n",
        "          'shufflenet_v2_x0_5',\n",
        "          'shufflenet_v2_x1_0',\n",
        "          'shufflenet_v2_x1_5',\n",
        "          'shufflenet_v2_x2_0',\n",
        "          'wide_resnet50_2',\n",
        "          'wide_resnet101_2']\n",
        "\n",
        "    model_string = 'global model; model = models.' + model_name + '(weights=\\'' + weights + '\\')'\n",
        "    exec(model_string)\n",
        "\n",
        "    model.name = model_name\n",
        "\n",
        "    if pretrained:\n",
        "        print('[INFO]: Loading pre-trained weights...')\n",
        "    else:\n",
        "        print('[INFO]: Not loading pre-trained weights.')\n",
        "    \n",
        "    if fine_tune:\n",
        "        print('[INFO]: Fine-tuning all layers...')\n",
        "        for params in model.parameters():\n",
        "            params.requires_grad = True\n",
        "    elif not fine_tune:\n",
        "        print('[INFO]: Freezing hidden layers...')\n",
        "        for params in model.parameters():\n",
        "            params.requires_grad = False\n",
        "    \n",
        "    if(model_name in fc):\n",
        "      num_ftrs = model.fc.in_features\n",
        "      model.fc = nn.Linear(in_features = num_ftrs, out_features = num_classes)\n",
        "    elif(model_name == 'inception_v3'):\n",
        "      aux_features = model.AuxLogits.fc.in_features\n",
        "      fc_features = model.fc.in_features\n",
        "      model.AuxLogits.fc = nn.Linear(aux_features, num_classes)\n",
        "      model.fc = nn.Linear(fc_features, num_classes)\n",
        "    elif(model_name == 'squeezenet1_0' or model_name == 'squeezenet1_1'):\n",
        "      in_chn = model.classifier[1].in_channels\n",
        "      model.classifier[1] = nn.Conv2d(in_channels = in_chn, out_channels = num_classes, kernel_size = (1,1), stride = (1,1))\n",
        "    elif(model_name == 'swin_t' or model_name == 'swin_s' or model_name == 'swin_b'):\n",
        "      num_ftrs = model.head.in_features\n",
        "      model.head = nn.Linear(in_features = num_ftrs, out_features = num_classes)\n",
        "    elif(model_name == 'vit_b_16' or model_name == 'vit_b_32' or model_name == 'vit_l_16' or model_name == 'vit_l_32' or model_name == 'vit_h_14'):\n",
        "      num_ftrs = model.heads.head.in_features\n",
        "      model.heads.head = nn.Linear(in_features = num_ftrs, out_features = num_classes)\n",
        "    else:\n",
        "      last_layer = model.classifier[-1]\n",
        "      num_ftrs = last_layer.in_features\n",
        "      model.classifier[-1] = nn.Linear(in_features = num_ftrs, out_features = num_classes)\n",
        "\n",
        "    return model\n",
        "\n",
        "\n",
        "\n",
        "model = build_model(model_name = 'resnet50', weights = 'DEFAULT', pretrained = True, fine_tune = True, num_classes = 4)\n",
        "\n",
        "model = model.to(device)\n",
        "\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "#criterion = nn.BCEWithLogitsLoss()\n",
        "\n",
        "optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)\n",
        "\n",
        "exp_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Import fine-tuned pretrained model\n",
        "\n",
        "model = torch.load('tumor_classification_model.pt')\n",
        "\n",
        "model = model.to(device)\n",
        "\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "\n",
        "optimizer_ft = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)\n",
        "\n",
        "exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=5, gamma=0.1)"
      ],
      "metadata": {
        "id": "xNVsp1F8EGGn"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Custom CNN Model\n",
        "\n",
        "class Net(nn.Module):\n",
        "\n",
        "    def __init__(self):\n",
        "        super(Net, self).__init__()\n",
        "        self.conv1 = nn.Conv2d(3, 32, 5)\n",
        "        self.pool = nn.MaxPool2d(2,2)\n",
        "        self.conv2 = nn.Conv2d(32, 64, 5)\n",
        "        self.conv3 = nn.Conv2d(64, 128, 5)\n",
        "        self.conv4 = nn.Conv2d(128, 256, 5)\n",
        "        self.fc1 = nn.Linear(256*12*12, 256)\n",
        "        self.fc2 = nn.Linear(256, 128)\n",
        "        self.fc3 = nn.Linear(128, 4)\n",
        "\n",
        "    def forward(self, x):\n",
        "        \n",
        "        x = self.pool(F.relu(self.conv1(x)))\n",
        "        x = self.pool(F.relu(self.conv2(x)))\n",
        "        x = self.pool(F.relu(self.conv3(x)))\n",
        "        x = self.pool(F.relu(self.conv4(x)))\n",
        "        \n",
        "        x = torch.flatten(x, 1)\n",
        "        x = F.relu(self.fc1(x))\n",
        "        x = F.relu(self.fc2(x))\n",
        "        x = self.fc3(x)\n",
        "        return x\n",
        "\n",
        "\n",
        "model = Net().to(device)\n",
        "\n",
        "criterion = nn.CrossEntropyLoss()\n",
        "optimizer_ft = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)\n",
        "exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)"
      ],
      "metadata": {
        "id": "RQN8fE1MvAiW"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Empty GPU Cache\n",
        "torch.cuda.empty_cache()\n",
        "\n",
        "# Empty Swap\n",
        "!swapoff -a && swapon -a\n",
        "\n",
        "#torch.cuda.memory_summary(device=None, abbreviated=False)\n",
        "#torch.cuda.memory_stats(device=device)"
      ],
      "metadata": {
        "id": "UAVXL3XjyXW5"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Train the model\n",
        "train_model(model, criterion, optimizer, exp_lr_scheduler, num_epochs=40)"
      ],
      "metadata": {
        "id": "44DbofGJ_H5S",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "031718b0-0e06-4719-9178-dfaee3ef6730"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.65823\n",
            "[EPOCH]: Training loss: 1.146905597006337   Validation Loss: 0.8633953332901001\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 2/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.81646\n",
            "[EPOCH]: Training loss: 0.6080959944912557   Validation Loss: 0.5282436013221741\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 3/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.88291\n",
            "[EPOCH]: Training loss: 0.3507322799623682   Validation Loss: 0.33691510558128357\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 4/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.89557\n",
            "[EPOCH]: Training loss: 0.2275897916615679   Validation Loss: 0.32370561361312866\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 5/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.91456\n",
            "[EPOCH]: Training loss: 0.16081062218781267   Validation Loss: 0.266632616519928\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 6/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.92722\n",
            "[EPOCH]: Training loss: 0.1184790562419744   Validation Loss: 0.23022495210170746\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 7/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.07810800338394186   Validation Loss: 0.2704782485961914\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 8/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.0710771608250111   Validation Loss: 0.19881270825862885\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 9/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.06201032732268063   Validation Loss: 0.26125556230545044\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 10/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.93987\n",
            "[EPOCH]: Training loss: 0.050001025560908435   Validation Loss: 0.18800349533557892\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 11/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.03793551938692003   Validation Loss: 0.22113949060440063\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 12/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.03302648383862433   Validation Loss: 0.1870686262845993\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 13/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.027702130965909427   Validation Loss: 0.20128564536571503\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 14/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.9462\n",
            "[EPOCH]: Training loss: 0.01607499452925214   Validation Loss: 0.17701146006584167\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 15/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.019646281515133014   Validation Loss: 0.18942970037460327\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 16/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.95253\n",
            "[EPOCH]: Training loss: 0.01711394699889476   Validation Loss: 0.1523049920797348\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 17/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.012850940328881438   Validation Loss: 0.16344188153743744\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 18/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.014064096760664009   Validation Loss: 0.15596938133239746\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 19/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.008837594547350839   Validation Loss: 0.158136785030365\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 20/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.010840937229046032   Validation Loss: 0.1923634558916092\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 21/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.01224791100235614   Validation Loss: 0.1563805341720581\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 22/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.011135810846201239   Validation Loss: 0.18165536224842072\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 23/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.009327589234682533   Validation Loss: 0.17103871703147888\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 24/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.012662956644807094   Validation Loss: 0.19965976476669312\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 25/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.00887789666813568   Validation Loss: 0.19449250400066376\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 26/40\n",
            "--------------------\n",
            "[INFO]: Model saved. Accuracy: 0.9557\n",
            "[EPOCH]: Training loss: 0.012444431224139407   Validation Loss: 0.1592879593372345\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 27/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.008062111780183453   Validation Loss: 0.17959555983543396\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 28/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.005045883432410009   Validation Loss: 0.17073577642440796\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 29/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.0049987771140223135   Validation Loss: 0.19786706566810608\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 30/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.004841785708879215   Validation Loss: 0.17533260583877563\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 31/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.00904004325013702   Validation Loss: 0.17605248093605042\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 32/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.004392855635548555   Validation Loss: 0.19569255411624908\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 33/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.004276159427094723   Validation Loss: 0.19482718408107758\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 34/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.003824543560726903   Validation Loss: 0.19153909385204315\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 35/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.00398277697314278   Validation Loss: 0.19095450639724731\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 36/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.004573717019932731   Validation Loss: 0.23105227947235107\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 37/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.0032297413424376836   Validation Loss: 0.19000957906246185\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 38/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.003290679237773914   Validation Loss: 0.21329575777053833\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 39/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.003582943799700117   Validation Loss: 0.18936891853809357\n",
            "------------------------------------------------------------------------------------------\n",
            "Epoch 40/40\n",
            "--------------------\n",
            "[EPOCH]: Training loss: 0.004884493418215784   Validation Loss: 0.2233033925294876\n",
            "------------------------------------------------------------------------------------------\n",
            "[TRAINING]: Training complete in 18m 41s\n",
            "[TRAINING]: Best val Accuracy: 0.955700\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 720x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Save the model\n",
        "torch.save(model,'tumor_classification_model.pt')"
      ],
      "metadata": {
        "id": "z8QEhWWMPkLV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Load and Test the model\n",
        "PATH = 'tc_resnet50.pt'\n",
        "\n",
        "if torch.cuda.is_available():\n",
        "  model = torch.load(PATH)\n",
        "  model.to(device)\n",
        "else:\n",
        "  model = torch.load(PATH, map_location=device)\n",
        "\n",
        "model.eval()\n",
        "\n",
        "test = test_model(model,validationloader)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 525
        },
        "id": "oZszVf10H-wb",
        "outputId": "cbdae9de-6664-4c94-bb7b-9c60bef0582e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "                  precision    recall  f1-score   support\n",
            "\n",
            "    glioma_tumor       0.98      0.93      0.95       108\n",
            "meningioma_tumor       0.91      0.94      0.93        86\n",
            "        no_tumor       0.92      0.95      0.94        38\n",
            " pituitary_tumor       0.98      1.00      0.99        84\n",
            "\n",
            "        accuracy                           0.95       316\n",
            "       macro avg       0.95      0.95      0.95       316\n",
            "    weighted avg       0.95      0.95      0.95       316\n",
            "\n",
            "heatmap_resnet50.png\n",
            "Accuracy: 0.95253\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "dFeYU1H5pU4k"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "tumor_classification.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyOH9eQt1W1MBTGzasepelbj"
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}