{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "machine_shape": "hm",
      "gpuType": "V100"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wS_st3blv2jp",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "193f0747-4cc9-4078-ded9-41b0b4a2a8ff"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1mqCsj6uO4WIbv3FRO9OTKXc8Wd8lYbKa\n",
            "To: /content/Dataset.zip\n",
            "\r  0%|          | 0.00/700M [00:00<?, ?B/s]\r  1%|          | 4.72M/700M [00:00<00:53, 12.9MB/s]\r  4%|▎         | 25.7M/700M [00:00<00:14, 46.8MB/s]\r  6%|▌         | 42.5M/700M [00:00<00:13, 50.5MB/s]\r  8%|▊         | 59.2M/700M [00:01<00:11, 56.1MB/s]\r 11%|█         | 76.0M/700M [00:01<00:10, 59.3MB/s]\r 13%|█▎        | 92.8M/700M [00:01<00:12, 48.8MB/s]\r 17%|█▋        | 118M/700M [00:02<00:09, 61.6MB/s] \r 18%|█▊        | 126M/700M [00:02<00:10, 57.1MB/s]\r 20%|██        | 143M/700M [00:02<00:07, 69.9MB/s]\r 22%|██▏       | 152M/700M [00:02<00:08, 68.1MB/s]\r 23%|██▎       | 160M/700M [00:02<00:09, 58.0MB/s]\r 25%|██▌       | 177M/700M [00:02<00:07, 71.4MB/s]\r 26%|██▋       | 185M/700M [00:03<00:08, 57.9MB/s]\r 28%|██▊       | 199M/700M [00:03<00:06, 72.2MB/s]\r 30%|███       | 211M/700M [00:03<00:05, 81.4MB/s]\r 32%|███▏      | 225M/700M [00:03<00:05, 93.4MB/s]\r 35%|███▍      | 244M/700M [00:03<00:03, 115MB/s] \r 37%|███▋      | 257M/700M [00:03<00:03, 119MB/s]\r 39%|███▊      | 270M/700M [00:03<00:03, 112MB/s]\r 42%|████▏     | 294M/700M [00:04<00:03, 128MB/s]\r 46%|████▌     | 319M/700M [00:04<00:02, 144MB/s]\r 48%|████▊     | 336M/700M [00:04<00:02, 142MB/s]\r 50%|█████     | 351M/700M [00:04<00:02, 129MB/s]\r 53%|█████▎    | 373M/700M [00:04<00:02, 152MB/s]\r 56%|█████▌    | 390M/700M [00:04<00:02, 145MB/s]\r 58%|█████▊    | 405M/700M [00:04<00:02, 127MB/s]\r 60%|██████    | 420M/700M [00:05<00:03, 92.7MB/s]\r 64%|██████▎   | 445M/700M [00:05<00:02, 112MB/s] \r 66%|██████▌   | 458M/700M [00:05<00:02, 111MB/s]\r 67%|██████▋   | 471M/700M [00:05<00:01, 115MB/s]\r 70%|██████▉   | 487M/700M [00:05<00:03, 66.5MB/s]\r 73%|███████▎  | 512M/700M [00:06<00:02, 90.7MB/s]\r 75%|███████▌  | 525M/700M [00:06<00:01, 88.7MB/s]\r 77%|███████▋  | 539M/700M [00:06<00:01, 97.1MB/s]\r 79%|███████▉  | 554M/700M [00:06<00:01, 99.8MB/s]\r 82%|████████▏ | 571M/700M [00:06<00:01, 108MB/s] \r 83%|████████▎ | 584M/700M [00:06<00:01, 85.2MB/s]\r 85%|████████▌ | 596M/700M [00:06<00:01, 91.2MB/s]\r 87%|████████▋ | 607M/700M [00:07<00:01, 84.1MB/s]\r 88%|████████▊ | 616M/700M [00:07<00:01, 68.3MB/s]\r 91%|█████████ | 634M/700M [00:07<00:00, 90.6MB/s]\r 92%|█████████▏| 645M/700M [00:07<00:00, 73.6MB/s]\r 94%|█████████▎| 655M/700M [00:07<00:00, 65.7MB/s]\r 96%|█████████▌| 672M/700M [00:08<00:00, 60.2MB/s]\r 99%|█████████▉| 691M/700M [00:08<00:00, 81.8MB/s]\r100%|██████████| 700M/700M [00:08<00:00, 84.1MB/s]\n",
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1CzMKzBpTBT_rhk-iMQAGe4_1Hsf5kmnE\n",
            "To: /content/TestDataSet.zip\n",
            "\r  0%|          | 0.00/833M [00:00<?, ?B/s]\r  1%|          | 4.72M/833M [00:00<01:24, 9.82MB/s]\r  3%|▎         | 25.7M/833M [00:00<00:27, 29.1MB/s]\r  5%|▌         | 42.5M/833M [00:01<00:24, 31.8MB/s]\r  7%|▋         | 58.2M/833M [00:01<00:16, 46.7MB/s]\r  8%|▊         | 66.6M/833M [00:01<00:20, 36.7MB/s]\r  9%|▉         | 76.0M/833M [00:02<00:24, 30.4MB/s]\r 11%|█▏        | 95.4M/833M [00:02<00:15, 48.6MB/s]\r 13%|█▎        | 105M/833M [00:02<00:18, 38.6MB/s] \r 15%|█▌        | 126M/833M [00:03<00:17, 40.7MB/s]\r 18%|█▊        | 152M/833M [00:03<00:15, 44.6MB/s]\r 21%|██        | 177M/833M [00:04<00:12, 54.5MB/s]\r 23%|██▎       | 193M/833M [00:04<00:12, 50.1MB/s]\r 25%|██▌       | 210M/833M [00:04<00:10, 61.9MB/s]\r 26%|██▋       | 220M/833M [00:05<00:16, 38.3MB/s]\r 29%|██▉       | 244M/833M [00:05<00:14, 41.7MB/s]\r 32%|███▏      | 269M/833M [00:06<00:12, 45.2MB/s]\r 36%|███▌      | 299M/833M [00:06<00:07, 67.7MB/s]\r 38%|███▊      | 313M/833M [00:06<00:09, 54.1MB/s]\r 39%|███▉      | 328M/833M [00:07<00:08, 60.4MB/s]\r 41%|████      | 338M/833M [00:07<00:09, 51.3MB/s]\r 43%|████▎     | 361M/833M [00:07<00:06, 68.4MB/s]\r 45%|████▍     | 372M/833M [00:07<00:08, 54.6MB/s]\r 47%|████▋     | 395M/833M [00:08<00:06, 70.8MB/s]\r 49%|████▉     | 410M/833M [00:08<00:05, 82.5MB/s]\r 51%|█████▏    | 428M/833M [00:08<00:04, 98.9MB/s]\r 53%|█████▎    | 445M/833M [00:08<00:03, 111MB/s] \r 55%|█████▌    | 459M/833M [00:08<00:03, 107MB/s]\r 57%|█████▋    | 478M/833M [00:08<00:02, 126MB/s]\r 59%|█████▉    | 493M/833M [00:08<00:02, 132MB/s]\r 61%|██████▏   | 511M/833M [00:08<00:02, 142MB/s]\r 63%|██████▎   | 528M/833M [00:08<00:02, 151MB/s]\r 65%|██████▌   | 545M/833M [00:09<00:02, 129MB/s]\r 69%|██████▊   | 571M/833M [00:09<00:02, 109MB/s]\r 72%|███████▏  | 600M/833M [00:09<00:01, 144MB/s]\r 74%|███████▍  | 618M/833M [00:09<00:02, 90.6MB/s]\r 76%|███████▌  | 631M/833M [00:09<00:02, 97.7MB/s]\r 78%|███████▊  | 646M/833M [00:10<00:01, 104MB/s] \r 79%|███████▉  | 660M/833M [00:10<00:01, 94.1MB/s]\r 81%|████████  | 674M/833M [00:10<00:01, 102MB/s] \r 83%|████████▎ | 688M/833M [00:10<00:01, 107MB/s]\r 84%|████████▍ | 701M/833M [00:10<00:01, 107MB/s]\r 86%|████████▌ | 713M/833M [00:10<00:01, 96.8MB/s]\r 88%|████████▊ | 730M/833M [00:10<00:00, 113MB/s] \r 89%|████████▉ | 743M/833M [00:11<00:00, 111MB/s]\r 91%|█████████ | 757M/833M [00:11<00:00, 118MB/s]\r 93%|█████████▎| 772M/833M [00:11<00:00, 105MB/s]\r 95%|█████████▍| 789M/833M [00:11<00:00, 119MB/s]\r 96%|█████████▋| 802M/833M [00:11<00:00, 101MB/s]\r100%|█████████▉| 831M/833M [00:11<00:00, 125MB/s]\r100%|██████████| 833M/833M [00:11<00:00, 70.9MB/s]\n"
          ]
        }
      ],
      "source": [
        "#@title 데이터셋 불러오기\n",
        "%%bash\n",
        "gdown 1mqCsj6uO4WIbv3FRO9OTKXc8Wd8lYbKa\n",
        "unzip -q Dataset.zip\n",
        "mkdir Dataset\n",
        "mv /content/masked /content/half /content/unmasked /content/Dataset\n",
        "gdown 1CzMKzBpTBT_rhk-iMQAGe4_1Hsf5kmnE\n",
        "unzip -q TestDataSet.zip"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision import transforms\n",
        "from torchvision.transforms import v2\n",
        "from torchvision.datasets import ImageFolder\n",
        "from sklearn.metrics import accuracy_score, f1_score\n",
        "from imblearn.over_sampling import SMOTE\n",
        "import numpy as np"
      ],
      "metadata": {
        "id": "xvPS6Xwcv6FM"
      },
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#@title half를 masked에 합치기\n",
        "import os\n",
        "import shutil\n",
        "half_folder = '/content/Dataset/half'  # \"half\" 폴더의 경로\n",
        "unmasked_folder = '/content/Dataset/masked'  # \"masked\" 폴더의 경로\n",
        "\n",
        "if not os.path.exists(unmasked_folder):\n",
        "    os.makedirs(unmasked_folder)\n",
        "for filename in os.listdir(half_folder):\n",
        "    source_path = os.path.join(half_folder, filename)\n",
        "    destination_path = os.path.join(unmasked_folder, filename)\n",
        "    if os.path.exists(destination_path):\n",
        "        new_filename = filename\n",
        "        name, ext = os.path.splitext(filename)\n",
        "        count = 1\n",
        "        while os.path.exists(os.path.join(unmasked_folder, new_filename)):\n",
        "            new_filename = f\"{name}_{count}{ext}\"\n",
        "            count += 1\n",
        "        destination_path = os.path.join(unmasked_folder, new_filename)\n",
        "    shutil.move(source_path, destination_path)\n",
        "shutil.rmtree(half_folder)\n",
        "print(\"파일 이동 및 폴더 삭제가 완료되었습니다.\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TsbVuJPkZShv",
        "outputId": "69ff0c72-38b3-4b53-d2df-b06c9a1c401a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "파일 이동 및 폴더 삭제가 완료되었습니다.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 이미지 데이터 폴더 경로 설정\n",
        "train_folder = '/content/Dataset'\n",
        "valid_folder = '/content/valid'\n",
        "\n",
        "# 이미지 전처리 및 데이터 로드\n",
        "transform = transforms.v2.Compose([\n",
        "    v2.Resize((224, 224)),\n",
        "    v2.RandomVerticalFlip(p=0.5),\n",
        "    #v2.AugMix(),\n",
        "    v2.Compose([v2.ToImage(), v2.ToDtype(torch.float32, scale=True)])\n",
        "])"
      ],
      "metadata": {
        "id": "i2QMdOq0wI17"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 훈련 데이터셋 생성\n",
        "train_dataset = ImageFolder(root=train_folder, transform=transform)\n",
        "# 검증 데이터셋 생성\n",
        "valid_dataset = ImageFolder(root=valid_folder, transform=transform)\n",
        "\n",
        "# DataLoader를 사용하여 데이터 배치로 로드 (배치 크기 변경)\n",
        "batch_size = 2048\n",
        "train_loader = DataLoader(train_dataset, num_workers=8, batch_size=batch_size, shuffle=True, pin_memory=True)\n",
        "valid_loader = DataLoader(valid_dataset, num_workers=8, batch_size=batch_size, shuffle=False, pin_memory=True)\n",
        "\n",
        "# 데이터 확인 (예시로 첫 번째 배치만 출력)\n",
        "#for batch in valid_loader:\n",
        "#    X_validation, y_validation = batch\n",
        "#    break"
      ],
      "metadata": {
        "id": "ldA26_nukhkl"
      },
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 데이터 일부만 로드하여 SMOTE 적용\n",
        "num_samples_for_smote = 1000  # 적절한 크기로 조절\n",
        "X_train = []\n",
        "y_train = []\n",
        "\n",
        "for batch in train_loader:\n",
        "    x_batch, y_batch = batch\n",
        "    X_train.append(x_batch.numpy())\n",
        "    y_train.append(y_batch.numpy())\n",
        "    if len(X_train) * batch_size >= num_samples_for_smote:\n",
        "        break\n",
        "\n",
        "X_train = np.concatenate(X_train)\n",
        "y_train = np.concatenate(y_train)\n",
        "\n",
        "# SMOTE 적용\n",
        "smote = SMOTE(sampling_strategy='auto', random_state=42)\n",
        "X_train_resampled, y_train_resampled = smote.fit_resample(X_train.reshape(-1, 224 * 224 * 3), y_train)\n",
        "\n",
        "# 텐서로 변환\n",
        "X_train_resampled = torch.tensor(X_train_resampled.reshape(-1, 3, 224, 224))\n",
        "y_train_resampled = torch.tensor(y_train_resampled)"
      ],
      "metadata": {
        "id": "CmUPTYdYkmDe"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "  # 로지스틱 회귀 모델 정의\n",
        "class LogisticRegressionModel(nn.Module):\n",
        "    def __init__(self, input_size):\n",
        "        super(LogisticRegressionModel, self).__init__()\n",
        "        self.flatten = nn.Flatten()\n",
        "        self.linear = nn.Linear(input_size, 1)\n",
        "        self.sigmoid = nn.Sigmoid()\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.flatten(x)\n",
        "        x = self.linear(x)\n",
        "        x = self.sigmoid(x)\n",
        "        return x"
      ],
      "metadata": {
        "id": "V5c8SK718-3W"
      },
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 모델 초기화\n",
        "input_size = 224 * 224 * 3  # 이미지 크기 및 채널 수에 따라 조정\n",
        "model = LogisticRegressionModel(input_size).cuda()\n",
        "\n",
        "# 손실 함수 및 최적화 알고리즘 설정 (AdamW)\n",
        "criterion = nn.BCELoss()\n",
        "optimizer = optim.AdamW(model.parameters(), lr=0.0001)\n"
      ],
      "metadata": {
        "id": "LDscXKbR8_4Q"
      },
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 학습\n",
        "num_epochs = 50\n",
        "for epoch in range(num_epochs):\n",
        "    model.train()\n",
        "    for i, (inputs, labels) in enumerate(train_loader):\n",
        "        inputs, labels = inputs.cuda(), labels.cuda()\n",
        "        optimizer.zero_grad()\n",
        "        outputs = model(inputs)\n",
        "\n",
        "        loss = criterion(outputs, labels.float().view(-1, 1))\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        total_steps = int(129600/batch_size)\n",
        "        print(f'Epoch:[{epoch+1}/{num_epochs}] Step:[{i+1}/{total_steps+1}]  Step_loss:{loss.item():.5f}')\n",
        "    # Print progress\n",
        "    print(f'Epoch [{epoch+1}/{num_epochs}], Loss: {loss.item()}')\n",
        "    # 모델 평가\n",
        "    model.eval()\n",
        "    all_predictions = []\n",
        "    all_labels = []\n",
        "    with torch.no_grad():\n",
        "      for inputs, labels in valid_loader:\n",
        "        inputs, labels = inputs.cuda(), labels.cuda()\n",
        "        outputs = model(inputs).cuda()\n",
        "        predictions = (outputs > 0.5).float()\n",
        "    all_predictions.append(predictions)\n",
        "    all_labels.append(labels)\n",
        "    all_predictions = torch.cat(all_predictions).cpu()\n",
        "    all_labels = torch.cat(all_labels).cpu()\n",
        "    accuracy = accuracy_score(all_labels, all_predictions)\n",
        "    print(f'Validation Accuracy: {accuracy}')\n",
        "    f1 = f1_score(all_labels, all_predictions)\n",
        "    print(f'Validation F1 Score: {f1}')"
      ],
      "metadata": {
        "id": "74dutkal9DtG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "00e0dfc4-22cd-4267-d47a-a858a895f6aa"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch [1/50] Step:[1/64]  Step_loss: 0.58745\n",
            "Epoch [1/50] Step:[2/64]  Step_loss: 0.58196\n",
            "Epoch [1/50] Step:[3/64]  Step_loss: 0.57389\n",
            "Epoch [1/50] Step:[4/64]  Step_loss: 0.59231\n",
            "Epoch [1/50] Step:[5/64]  Step_loss: 0.57840\n",
            "Epoch [1/50] Step:[6/64]  Step_loss: 0.57138\n",
            "Epoch [1/50] Step:[7/64]  Step_loss: 0.57638\n",
            "Epoch [1/50] Step:[8/64]  Step_loss: 0.56401\n",
            "Epoch [1/50] Step:[9/64]  Step_loss: 0.58184\n",
            "Epoch [1/50] Step:[10/64]  Step_loss: 0.57189\n",
            "Epoch [1/50] Step:[11/64]  Step_loss: 0.56412\n",
            "Epoch [1/50] Step:[12/64]  Step_loss: 0.57424\n",
            "Epoch [1/50] Step:[13/64]  Step_loss: 0.56927\n",
            "Epoch [1/50] Step:[14/64]  Step_loss: 0.55216\n",
            "Epoch [1/50] Step:[15/64]  Step_loss: 0.57783\n",
            "Epoch [1/50] Step:[16/64]  Step_loss: 0.55716\n",
            "Epoch [1/50] Step:[17/64]  Step_loss: 0.56030\n",
            "Epoch [1/50] Step:[18/64]  Step_loss: 0.56626\n",
            "Epoch [1/50] Step:[19/64]  Step_loss: 0.54872\n",
            "Epoch [1/50] Step:[20/64]  Step_loss: 0.56241\n",
            "Epoch [1/50] Step:[21/64]  Step_loss: 0.56201\n",
            "Epoch [1/50] Step:[22/64]  Step_loss: 0.55821\n",
            "Epoch [1/50] Step:[23/64]  Step_loss: 0.56016\n",
            "Epoch [1/50] Step:[24/64]  Step_loss: 0.56836\n",
            "Epoch [1/50] Step:[25/64]  Step_loss: 0.55823\n",
            "Epoch [1/50] Step:[26/64]  Step_loss: 0.55694\n",
            "Epoch [1/50] Step:[27/64]  Step_loss: 0.55958\n",
            "Epoch [1/50] Step:[28/64]  Step_loss: 0.55113\n",
            "Epoch [1/50] Step:[29/64]  Step_loss: 0.55596\n",
            "Epoch [1/50] Step:[30/64]  Step_loss: 0.54465\n",
            "Epoch [1/50] Step:[31/64]  Step_loss: 0.54343\n",
            "Epoch [1/50] Step:[32/64]  Step_loss: 0.54366\n",
            "Epoch [1/50] Step:[33/64]  Step_loss: 0.54128\n",
            "Epoch [1/50] Step:[34/64]  Step_loss: 0.54744\n",
            "Epoch [1/50] Step:[35/64]  Step_loss: 0.55186\n",
            "Epoch [1/50] Step:[36/64]  Step_loss: 0.54697\n",
            "Epoch [1/50] Step:[37/64]  Step_loss: 0.54832\n",
            "Epoch [1/50] Step:[38/64]  Step_loss: 0.54394\n",
            "Epoch [1/50] Step:[39/64]  Step_loss: 0.53079\n",
            "Epoch [1/50] Step:[40/64]  Step_loss: 0.54948\n",
            "Epoch [1/50] Step:[41/64]  Step_loss: 0.53726\n",
            "Epoch [1/50] Step:[42/64]  Step_loss: 0.54393\n",
            "Epoch [1/50] Step:[43/64]  Step_loss: 0.53918\n",
            "Epoch [1/50] Step:[44/64]  Step_loss: 0.54003\n",
            "Epoch [1/50] Step:[45/64]  Step_loss: 0.54803\n",
            "Epoch [1/50] Step:[46/64]  Step_loss: 0.53045\n",
            "Epoch [1/50] Step:[47/64]  Step_loss: 0.53130\n",
            "Epoch [1/50] Step:[48/64]  Step_loss: 0.53213\n",
            "Epoch [1/50] Step:[49/64]  Step_loss: 0.53309\n",
            "Epoch [1/50] Step:[50/64]  Step_loss: 0.53307\n",
            "Epoch [1/50] Step:[51/64]  Step_loss: 0.53282\n",
            "Epoch [1/50] Step:[52/64]  Step_loss: 0.52812\n",
            "Epoch [1/50] Step:[53/64]  Step_loss: 0.52012\n",
            "Epoch [1/50] Step:[54/64]  Step_loss: 0.52923\n",
            "Epoch [1/50] Step:[55/64]  Step_loss: 0.52933\n",
            "Epoch [1/50] Step:[56/64]  Step_loss: 0.52348\n",
            "Epoch [1/50] Step:[57/64]  Step_loss: 0.52567\n",
            "Epoch [1/50] Step:[58/64]  Step_loss: 0.51367\n",
            "Epoch [1/50] Step:[59/64]  Step_loss: 0.53834\n",
            "Epoch [1/50] Step:[60/64]  Step_loss: 0.52386\n",
            "Epoch [1/50] Step:[61/64]  Step_loss: 0.53638\n",
            "Epoch [1/50] Step:[62/64]  Step_loss: 0.51488\n",
            "Epoch [1/50] Step:[63/64]  Step_loss: 0.54372\n",
            "Epoch [1/50] Step:[64/64]  Step_loss: 0.53531\n",
            "Epoch [1/50], Loss: 0.5353068709373474\n",
            "Validation Accuracy: 0.6314377682403434\n",
            "Validation F1 Score: 0.7740874712265702\n",
            "Epoch [2/50] Step:[1/64]  Step_loss: 0.53095\n",
            "Epoch [2/50] Step:[2/64]  Step_loss: 0.52171\n",
            "Epoch [2/50] Step:[3/64]  Step_loss: 0.53188\n",
            "Epoch [2/50] Step:[4/64]  Step_loss: 0.50675\n",
            "Epoch [2/50] Step:[5/64]  Step_loss: 0.51028\n",
            "Epoch [2/50] Step:[6/64]  Step_loss: 0.52670\n",
            "Epoch [2/50] Step:[7/64]  Step_loss: 0.51836\n",
            "Epoch [2/50] Step:[8/64]  Step_loss: 0.51599\n",
            "Epoch [2/50] Step:[9/64]  Step_loss: 0.50562\n",
            "Epoch [2/50] Step:[10/64]  Step_loss: 0.49185\n",
            "Epoch [2/50] Step:[11/64]  Step_loss: 0.51065\n",
            "Epoch [2/50] Step:[12/64]  Step_loss: 0.50994\n",
            "Epoch [2/50] Step:[13/64]  Step_loss: 0.50348\n",
            "Epoch [2/50] Step:[14/64]  Step_loss: 0.50455\n",
            "Epoch [2/50] Step:[15/64]  Step_loss: 0.49326\n",
            "Epoch [2/50] Step:[16/64]  Step_loss: 0.50965\n",
            "Epoch [2/50] Step:[17/64]  Step_loss: 0.50516\n",
            "Epoch [2/50] Step:[18/64]  Step_loss: 0.50401\n",
            "Epoch [2/50] Step:[19/64]  Step_loss: 0.50499\n",
            "Epoch [2/50] Step:[20/64]  Step_loss: 0.49590\n",
            "Epoch [2/50] Step:[21/64]  Step_loss: 0.50605\n",
            "Epoch [2/50] Step:[22/64]  Step_loss: 0.50388\n",
            "Epoch [2/50] Step:[23/64]  Step_loss: 0.49169\n",
            "Epoch [2/50] Step:[24/64]  Step_loss: 0.50895\n",
            "Epoch [2/50] Step:[25/64]  Step_loss: 0.49507\n",
            "Epoch [2/50] Step:[26/64]  Step_loss: 0.49121\n",
            "Epoch [2/50] Step:[27/64]  Step_loss: 0.48667\n",
            "Epoch [2/50] Step:[28/64]  Step_loss: 0.50104\n",
            "Epoch [2/50] Step:[29/64]  Step_loss: 0.49185\n",
            "Epoch [2/50] Step:[30/64]  Step_loss: 0.50014\n",
            "Epoch [2/50] Step:[31/64]  Step_loss: 0.51080\n",
            "Epoch [2/50] Step:[32/64]  Step_loss: 0.48850\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 모델 평가\n",
        "model.eval()\n",
        "all_predictions = []\n",
        "all_labels = []\n",
        "with torch.no_grad():\n",
        "    for inputs, labels in valid_loader:\n",
        "        inputs, labels = inputs.cuda(), labels.cuda()\n",
        "        outputs = model(inputs).cuda()\n",
        "        predictions = (outputs > 0.5).float()\n",
        "        all_predictions.append(predictions)\n",
        "        all_labels.append(labels)\n",
        "\n",
        "# 전체 데이터에 대한 예측과 레이블을 하나로 합침\n",
        "all_predictions = torch.cat(all_predictions).cpu()\n",
        "all_labels = torch.cat(all_labels).cpu()"
      ],
      "metadata": {
        "id": "RLWn-yczZ4RY"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 정확도 계산\n",
        "accuracy = accuracy_score(all_labels, all_predictions)\n",
        "print(f'Validation Accuracy: {accuracy}')\n",
        "\n",
        "# F1 스코어 계산\n",
        "f1 = f1_score(all_labels, all_predictions)\n",
        "print(f'Validation F1 Score: {f1}')"
      ],
      "metadata": {
        "id": "H86915ck9E4a",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e45d2686-7660-4690-8c47-a19a881ed42c"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Validation Accuracy: 0.8342592592592593\n",
            "Validation F1 Score: 0.76177801437317\n"
          ]
        }
      ]
    }
  ]
}