| import os |
| import numpy as np |
| from PIL import Image |
| |
|
|
| INPUT_NPZ_PATH = r"./octmnist.npz" |
| OUTPUT_ROOT = r"" |
| |
| SAVE_IN_LABEL_FOLDERS = True |
| |
| DATASET_LABEL_MAPS = { |
| "pathmnist": { |
| 0: "adipose", |
| 1: "background", |
| 2: "debris", |
| 3: "lymphocytes", |
| 4: "mucus", |
| 5: "smooth muscle", |
| 6: "normal colon mucosa", |
| 7: "cancer-associated stroma", |
| 8: "colorectal adenocarcinoma epithelium", |
| }, |
| "chestmnist": { |
| 0: "atelectasis", |
| 1: "cardiomegaly", |
| 2: "effusion", |
| 3: "infiltration", |
| 4: "mass", |
| 5: "nodule", |
| 6: "pneumonia", |
| 7: "pneumothorax", |
| 8: "consolidation", |
| 9: "edema", |
| 10: "emphysema", |
| 11: "fibrosis", |
| 12: "pleural", |
| 13: "hernia", |
| }, |
| "dermamnist": { |
| 0: "actinic keratoses and intraepithelial carcinoma", |
| 1: "basal cell carcinoma", |
| 2: "benign keratosis-like lesions", |
| 3: "dermatofibroma", |
| 4: "melanoma", |
| 5: "melanocytic nevi", |
| 6: "vascular lesions", |
| }, |
| "octmnist": { |
| 0: "choroidal neovascularization", |
| 1: "diabetic macular edema", |
| 2: "drusen", |
| 3: "normal", |
| }, |
| "pneumoniamnist": { |
| 0: "normal", |
| 1: "pneumonia", |
| }, |
| "retinamnist": { |
| 0: "0", |
| 1: "1", |
| 2: "2", |
| 3: "3", |
| 4: "4", |
| }, |
| "breastmnist": { |
| 0: "malignant", |
| 1: "normal, benign", |
| }, |
| "bloodmnist": { |
| 0: "basophil", |
| 1: "eosinophil", |
| 2: "erythroblast", |
| 3: "immature granulocytes(myelocytes, metamyelocytes and promyelocytes)", |
| 4: "lymphocyte", |
| 5: "monocyte", |
| 6: "neutrophil", |
| 7: "platelet", |
| }, |
| "tissuemnist": { |
| 0: "Collecting Duct, Connecting Tubule", |
| 1: "Distal Convoluted Tubule", |
| 2: "Glomerular endothelial cells", |
| 3: "Interstitial endothelial cells", |
| 4: "Leukocytes", |
| 5: "Podocytes", |
| 6: "Proximal Tubule Segments", |
| 7: "Thick Ascending Limb", |
| }, |
| "organamnist": { |
| 0: "bladder", |
| 1: "femur-left", |
| 2: "femur-right", |
| 3: "heart", |
| 4: "kidney-left", |
| 5: "kidney-right", |
| 6: "liver", |
| 7: "lung-left", |
| 8: "lung-right", |
| 9: "pancreas", |
| 10: "spleen", |
| }, |
| "organcmnist": { |
| 0: "bladder", |
| 1: "femur-left", |
| 2: "femur-right", |
| 3: "heart", |
| 4: "kidney-left", |
| 5: "kidney-right", |
| 6: "liver", |
| 7: "lung-left", |
| 8: "lung-right", |
| 9: "pancreas", |
| 10: "spleen", |
| }, |
| "organsmnist": { |
| 0: "bladder", |
| 1: "femur-left", |
| 2: "femur-right", |
| 3: "heart", |
| 4: "kidney-left", |
| 5: "kidney-right", |
| 6: "liver", |
| 7: "lung-left", |
| 8: "lung-right", |
| 9: "pancreas", |
| 10: "spleen", |
| }, |
| "organmnist3d": { |
| 0: "liver", |
| 1: "kidney-right", |
| 2: "kidney-left", |
| 3: "femur-right", |
| 4: "femur-left", |
| 5: "bladder", |
| 6: "heart", |
| 7: "lung-right", |
| 8: "lung-left", |
| 9: "spleen", |
| 10: "pancreas", |
| }, |
| "nodulemnist3d": { |
| 0: "benign", |
| 1: "malignant", |
| }, |
| "adrenalmnist3d": { |
| 0: "normal", |
| 1: "hyperplasia", |
| }, |
| "fracturemnist3d": { |
| 0: "buckle rib fracture", |
| 1: "nondisplaced rib fracture", |
| 2: "displaced rib fracture", |
| }, |
| "vesselmnist3d": { |
| 0: "vessel", |
| 1: "aneurysm", |
| }, |
| "synapsemnist3d": { |
| 0: "inhibitory synapse", |
| 1: "excitatory synapse", |
| }, |
| } |
| |
|
|
| def resolve_label_map(npz_path): |
| """ |
| Extract the dataset name from the NPZ filename and return its label map. |
| Handles suffixes like _64, _128, _224 (e.g. 'tissuemnist_224.npz' -> 'tissuemnist'). |
| Returns None if the dataset is not recognized. |
| """ |
| basename = os.path.basename(npz_path) |
| stem = os.path.splitext(basename)[0] |
|
|
| for suffix in ("_28", "_64", "_128", "_224"): |
| if stem.endswith(suffix): |
| stem = stem[: -len(suffix)] |
| break |
| |
| dataset_key = stem.lower() |
| |
| if dataset_key in DATASET_LABEL_MAPS: |
| print(f"[Info] Detected dataset: '{dataset_key}' — label map loaded automatically.") |
| return DATASET_LABEL_MAPS[dataset_key] |
| else: |
| print(f"[Warning] Dataset '{dataset_key}' not found in label map registry. " |
| "Folders will be named by integer label.") |
| return None |
| |
| def ensure_dir(path): |
| os.makedirs(path, exist_ok=True) |
| |
| |
| def save_one_image(img_array, save_path): |
| img_array = np.asarray(img_array) |
| |
| if img_array.dtype != np.uint8: |
| img_array = img_array.astype(np.uint8) |
| |
| if img_array.ndim == 2: |
| image = Image.fromarray(img_array, mode="L") |
| elif img_array.ndim == 3: |
| if img_array.shape[2] == 1: |
| image = Image.fromarray(img_array[:, :, 0], mode="L") |
| elif img_array.shape[2] == 3: |
| image = Image.fromarray(img_array, mode="RGB") |
| elif img_array.shape[2] == 4: |
| image = Image.fromarray(img_array, mode="RGBA") |
| else: |
| raise ValueError(f"Unsupported number of channels: {img_array.shape}") |
| else: |
| raise ValueError(f"Unsupported image shape: {img_array.shape}") |
| |
| image.save(save_path) |
| |
| |
| def label_to_folder_name(label_value, label_map): |
| label_value = int(label_value) |
| if label_map is not None and label_value in label_map: |
| return label_map[label_value] |
| return str(label_value) |
| |
| |
| def process_split(images, labels, split_name, label_map): |
| split_root = os.path.join(OUTPUT_ROOT, split_name) |
| ensure_dir(split_root) |
| |
| total = len(images) |
| success = 0 |
| |
| for i in range(total): |
| img = images[i] |
| label = labels[i] |
| |
| |
| if isinstance(label, np.ndarray): |
| if label.size == 1: |
| label = int(label.reshape(-1)[0]) |
| else: |
| raise ValueError(f"[{split_name}] Unexpected label format at index {i}: {label}") |
| else: |
| label = int(label) |
| |
| if SAVE_IN_LABEL_FOLDERS: |
| class_name = label_to_folder_name(label, label_map) |
| save_dir = os.path.join(split_root, class_name) |
| else: |
| save_dir = split_root |
| |
| ensure_dir(save_dir) |
| |
| save_name = f"{split_name}_{i:06d}.png" |
| save_path = os.path.join(save_dir, save_name) |
| |
| try: |
| save_one_image(img, save_path) |
| success += 1 |
| except Exception as e: |
| print(f"[Skip] {split_name} image {i} could not be saved: {e}") |
| |
| print(f"[Done] {split_name}: saved {success} / {total} images successfully.") |
| |
|
|
| def main(): |
| ensure_dir(OUTPUT_ROOT) |
| |
| print(f"[Info] Reading NPZ file: {INPUT_NPZ_PATH}") |
| data = np.load(INPUT_NPZ_PATH) |
| |
| print("\n[Info] Fields found in NPZ file:") |
| for k in data.files: |
| arr = data[k] |
| print(f" - {k}: shape={arr.shape}, dtype={arr.dtype}") |
| |
| required_keys = [ |
| "train_images", "train_labels", |
| "val_images", "val_labels", |
| "test_images", "test_labels", |
| ] |
| for key in required_keys: |
| if key not in data: |
| raise KeyError(f"[Error] Missing required field: '{key}'") |
| |
| label_map = resolve_label_map(INPUT_NPZ_PATH) |
| |
| train_images, train_labels = data["train_images"], data["train_labels"] |
| val_images, val_labels = data["val_images"], data["val_labels"] |
| test_images, test_labels = data["test_images"], data["test_labels"] |
| |
| print("\n[Info] Extracting and saving images...") |
| process_split(train_images, train_labels, "train", label_map) |
| process_split(val_images, val_labels, "val", label_map) |
| process_split(test_images, test_labels, "test", label_map) |
| |
| print(f"\n[Info] All done. Output directory: {OUTPUT_ROOT}") |
| |
| |
| if __name__ == "__main__": |
| main() |
|
|
