{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.12.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceType":"competition","sourceId":10418,"databundleVersionId":862236}],"dockerImageVersionId":31328,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import os\n\nimport numpy as np\nimport pandas as pd\n\nimport einops\n\nimport cv2\nfrom PIL import Image\nimport albumentations as A\n\nimport torch\nimport torch.nn as nn\nfrom torch.utils.data import Dataset, DataLoader\nimport torchvision.transforms as T\nimport torch.nn.functional as F\n\nimport matplotlib.pyplot as plt\nimport seaborn as sns","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2026-05-04T23:47:44.051116Z","iopub.execute_input":"2026-05-04T23:47:44.051815Z","iopub.status.idle":"2026-05-04T23:47:54.321941Z","shell.execute_reply.started":"2026-05-04T23:47:44.051783Z","shell.execute_reply":"2026-05-04T23:47:54.321261Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"p = \"/kaggle/input/competitions/human-protein-atlas-image-classification/train/\"\nimages = []\n\nfor n in [\n    \"00070df0-bbc3-11e8-b2bc-ac1f6b6435d0\", \n    \"000a6c98-bb9b-11e8-b2b9-ac1f6b6435d0\", \n    \"000a9596-bbc4-11e8-b2bc-ac1f6b6435d0\", \n    \"000c99ba-bba4-11e8-b2b9-ac1f6b6435d0\"\n]:\n    image = [\n        cv2.imread(f\"{p}{n}_{c}.png\", cv2.IMREAD_GRAYSCALE).astype(np.float32)/255 for c in ['red','green','blue','yellow']\n    ]\n    images += [np.stack(image, axis=-1)]\n\nimages[0].shape","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-05-04T23:47:54.512007Z","iopub.execute_input":"2026-05-04T23:47:54.512339Z","iopub.status.idle":"2026-05-04T23:47:54.586697Z","shell.execute_reply.started":"2026-05-04T23:47:54.512314Z","shell.execute_reply":"2026-05-04T23:47:54.585848Z"},"jupyter":{"source_hidden":true}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def visualization(images):\n    \n    fig, axes = plt.subplots(1, 4, figsize=(20, 5))\n    \n    for i, image in enumerate(images):\n        \n        composite = np.zeros_like(image[:,:,:3])\n        \n        composite[:,:,0] = image[:,:,0] + image[:,:,3] * .5\n        composite[:,:,1] = image[:,:,1] + image[:,:,3] * .5\n        composite[:,:,2] = image[:,:,2]\n        \n        composite = np.clip(composite, 0, 1)\n        \n        axes[i].imshow(composite)\n        axes[i].axis('off')\n    \n    plt.tight_layout()\n    plt.show()\n\nvisualization(images)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-05-04T22:53:24.652996Z","iopub.execute_input":"2026-05-04T22:53:24.653287Z","iopub.status.idle":"2026-05-04T22:53:25.263594Z","shell.execute_reply.started":"2026-05-04T22:53:24.653260Z","shell.execute_reply":"2026-05-04T22:53:25.262289Z"},"collapsed":true,"jupyter":{"source_hidden":true,"outputs_hidden":true}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"transform = A.Compose([\n    A.HorizontalFlip(p=.5),\n    A.VerticalFlip(p=.3),\n    A.ShiftScaleRotate(\n        shift_limit=.1, \n        scale_limit=.2, \n        rotate_limit=30, \n        p=.5\n    )\n])\n\nvisualization([transform(image=image)['image'] for image in images])","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-05-04T22:55:00.543021Z","iopub.execute_input":"2026-05-04T22:55:00.543316Z","iopub.status.idle":"2026-05-04T22:55:01.112595Z","shell.execute_reply.started":"2026-05-04T22:55:00.543290Z","shell.execute_reply":"2026-05-04T22:55:01.111718Z"},"jupyter":{"source_hidden":true,"outputs_hidden":true},"collapsed":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"class ProteinAtlasDataset(Dataset):\n    def __init__(self, csv_file, root_dir, transform=None):\n        self.df = pd.read_csv(csv_file)\n        self.root_dir = root_dir\n        self.transform = transform\n        self.num_classes = 28\n\n    def __len__(self):\n        return len(self.df)\n\n    def __getitem__(self, idx):\n        if torch.is_tensor(idx):\n            idx = idx.tolist()\n\n        img_id = self.df.iloc[idx, 0]\n        label_str = str(self.df.iloc[idx, 1]).split()\n\n        channels = ['red', 'green', 'blue', 'yellow']\n        img_arrays = []\n        \n        for ch in channels:\n            img_path = os.path.join(self.root_dir, f\"{img_id}_{ch}.png\")\n            \n            img = Image.open(img_path).convert('L')\n            img_arrays.append(np.array(img))\n            \n        image_tensor = torch.tensor(np.stack(img_arrays), dtype=torch.float32) / 255.0\n        \n        target = torch.zeros(self.num_classes, dtype=torch.float32)\n        for label in label_str:\n            if label != 'nan':\n                target[int(label)] = 1.0\n                \n        if self.transform:\n            image_tensor = self.transform(image_tensor)\n            \n        return image_tensor, target","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-05-04T23:47:54.588096Z","iopub.execute_input":"2026-05-04T23:47:54.588453Z","iopub.status.idle":"2026-05-04T23:47:54.595273Z","shell.execute_reply.started":"2026-05-04T23:47:54.588427Z","shell.execute_reply":"2026-05-04T23:47:54.594393Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"train_transforms = T.Compose([\n    T.Resize((256, 256), antialias=True),\n    T.RandomHorizontalFlip(p=0.5),\n    T.RandomVerticalFlip(p=0.5),\n    T.RandomRotation(degrees=30),\n])\n\nval_transforms = T.Compose([\n    T.Resize((256, 256), antialias=True)\n])","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-05-04T23:47:58.585397Z","iopub.execute_input":"2026-05-04T23:47:58.585672Z","iopub.status.idle":"2026-05-04T23:47:58.590544Z","shell.execute_reply.started":"2026-05-04T23:47:58.585648Z","shell.execute_reply":"2026-05-04T23:47:58.589732Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"CSV_PATH = \"/kaggle/input/competitions/human-protein-atlas-image-classification/train.csv\"\nIMG_DIR = \"/kaggle/input/competitions/human-protein-atlas-image-classification/train/\"\n\ntrain_dataset = ProteinAtlasDataset(\n    csv_file=CSV_PATH, \n    root_dir=IMG_DIR, \n    transform=train_transforms\n)\n\ntrain_loader = DataLoader(\n    train_dataset,\n    batch_size=32,\n    shuffle=True,\n    num_workers=4,\n    pin_memory=True\n)\n\n\nif __name__ == \"__main__\":\n    images, labels = next(iter(train_loader))\n    \n    print(f\"Image batch shape: {images.shape}\")\n    print(f\"Label batch shape: {labels.shape}\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-05-04T23:48:27.612382Z","iopub.execute_input":"2026-05-04T23:48:27.613417Z","iopub.status.idle":"2026-05-04T23:48:31.325556Z","shell.execute_reply.started":"2026-05-04T23:48:27.613333Z","shell.execute_reply":"2026-05-04T23:48:31.324736Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}