| |
| """ |
| Prepare a one-class, one-modality MILK10k folder for Stable Diffusion training. |
| |
| MILK10k has two images per lesion/sample: |
| - clinical: close-up |
| - dermoscopic |
| |
| This script exports only one requested modality. For example, if you prepare |
| clinical_close_up MEL data, the paired dermoscopic MEL image from the same |
| lesion_id is not copied. |
| |
| The output format is simple and works with common DreamBooth/LoRA trainers: |
| <output-dir>/images/*.jpg |
| <output-dir>/images/*.txt |
| <output-dir>/training_manifest.csv |
| """ |
|
|
| from __future__ import annotations |
|
|
| import argparse |
| import csv |
| import random |
| import shutil |
| from pathlib import Path |
|
|
| LABEL_COLUMNS = [ |
| "AKIEC", |
| "BCC", |
| "BEN_OTH", |
| "BKL", |
| "DF", |
| "INF", |
| "MAL_OTH", |
| "MEL", |
| "NV", |
| "SCCKA", |
| "VASC", |
| ] |
|
|
| CLASS_PROMPTS = { |
| "AKIEC": "actinic keratosis or intraepithelial carcinoma skin lesion", |
| "BCC": "basal cell carcinoma skin lesion", |
| "BEN_OTH": "benign other skin lesion", |
| "BKL": "benign keratosis-like skin lesion", |
| "DF": "dermatofibroma skin lesion", |
| "INF": "inflammatory skin lesion", |
| "MAL_OTH": "other malignant skin lesion", |
| "MEL": "melanoma skin lesion", |
| "NV": "melanocytic nevus skin lesion", |
| "SCCKA": "squamous cell carcinoma or keratoacanthoma skin lesion", |
| "VASC": "vascular skin lesion", |
| } |
|
|
| IMAGE_TYPE_PROMPTS = { |
| "clinical_close_up": "clinical close-up dermatology photograph", |
| "dermoscopic": "dermoscopic dermatology image", |
| } |
|
|
|
|
| def parse_args() -> argparse.Namespace: |
| parser = argparse.ArgumentParser(description="Export MILK10k images for one-class Stable Diffusion training.") |
| parser.add_argument("--data-dir", type=Path, default=Path("."), help="MILK10k root folder.") |
| parser.add_argument("--output-dir", type=Path, required=True, help="Output training folder.") |
| parser.add_argument("--class-name", required=True, choices=LABEL_COLUMNS) |
| parser.add_argument( |
| "--image-type", |
| required=True, |
| choices=["clinical_close_up", "dermoscopic"], |
| help="Export exactly one modality. clinical_close_up never copies dermoscopic images, and vice versa.", |
| ) |
| parser.add_argument("--max-images", type=int, default=None) |
| parser.add_argument("--seed", type=int, default=42) |
| parser.add_argument("--shuffle", action="store_true") |
| parser.add_argument("--copy", action="store_true", help="Copy images instead of creating symlinks.") |
| parser.add_argument("--caption", default=None, help="Override caption text written next to each image.") |
| return parser.parse_args() |
|
|
|
|
| def normalize_image_type(image_type: str) -> str: |
| if image_type == "clinical: close-up": |
| return "clinical_close_up" |
| return image_type.replace(" ", "_").replace(":", "").replace("-", "_") |
|
|
|
|
| def load_class_modality_rows(data_dir: Path, class_name: str, image_type: str) -> list[dict[str, str]]: |
| input_dir = data_dir / "MILK10k_Training_Input" |
| gt_path = data_dir / "MILK10k_Training_GroundTruth.csv" |
| meta_path = data_dir / "MILK10k_Training_Metadata.csv" |
| if not gt_path.exists() or not meta_path.exists() or not input_dir.exists(): |
| raise FileNotFoundError( |
| "Expected MILK10k_Training_GroundTruth.csv, MILK10k_Training_Metadata.csv, " |
| "and MILK10k_Training_Input under --data-dir." |
| ) |
|
|
| selected_lesions = set() |
| with gt_path.open(newline="") as f: |
| for row in csv.DictReader(f): |
| if float(row[class_name]) == 1.0: |
| selected_lesions.add(row["lesion_id"]) |
|
|
| rows = [] |
| with meta_path.open(newline="") as f: |
| for row in csv.DictReader(f): |
| if row["lesion_id"] not in selected_lesions: |
| continue |
| image_type_norm = normalize_image_type(row["image_type"]) |
| if image_type_norm != image_type: |
| continue |
| source_path = input_dir / row["lesion_id"] / f"{row['isic_id']}.jpg" |
| if not source_path.exists(): |
| continue |
| rows.append( |
| { |
| "lesion_id": row["lesion_id"], |
| "isic_id": row["isic_id"], |
| "image_type_norm": image_type_norm, |
| "source_path": str(source_path), |
| } |
| ) |
|
|
| if not rows: |
| raise ValueError(f"No images found for class={class_name}, image_type={image_type}") |
| return rows |
|
|
|
|
| def link_or_copy(src: Path, dst: Path, copy_file: bool) -> None: |
| if dst.exists() or dst.is_symlink(): |
| dst.unlink() |
| if copy_file: |
| shutil.copy2(src, dst) |
| else: |
| dst.symlink_to(src.resolve()) |
|
|
|
|
| def main() -> None: |
| args = parse_args() |
| data_dir = args.data_dir.expanduser().resolve() |
| output_dir = args.output_dir.expanduser().resolve() |
| images_dir = output_dir / "images" |
| images_dir.mkdir(parents=True, exist_ok=True) |
|
|
| rows = load_class_modality_rows(data_dir, args.class_name, args.image_type) |
| if args.shuffle: |
| random.Random(args.seed).shuffle(rows) |
| if args.max_images is not None: |
| rows = rows[: args.max_images] |
|
|
| caption = args.caption or f"{CLASS_PROMPTS[args.class_name]}, {IMAGE_TYPE_PROMPTS[args.image_type]}" |
| manifest_path = output_dir / "training_manifest.csv" |
|
|
| with manifest_path.open("w", newline="") as f: |
| fieldnames = ["class_name", "image_type", "lesion_id", "isic_id", "source_path", "train_image_path", "caption"] |
| writer = csv.DictWriter(f, fieldnames=fieldnames) |
| writer.writeheader() |
|
|
| for row in rows: |
| dst_image = images_dir / f"{row['lesion_id']}_{row['isic_id']}_{args.image_type}.jpg" |
| dst_caption = dst_image.with_suffix(".txt") |
| link_or_copy(Path(row["source_path"]), dst_image, args.copy) |
| dst_caption.write_text(caption + "\n", encoding="utf-8") |
| writer.writerow( |
| { |
| "class_name": args.class_name, |
| "image_type": args.image_type, |
| "lesion_id": row["lesion_id"], |
| "isic_id": row["isic_id"], |
| "source_path": row["source_path"], |
| "train_image_path": str(dst_image), |
| "caption": caption, |
| } |
| ) |
|
|
| print(f"Exported {len(rows)} {args.image_type} images for class {args.class_name}") |
| print(f"Images/captions: {images_dir}") |
| print(f"Manifest: {manifest_path}") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
