clean_dataset.py

import csv
import re
import shutil
from pathlib import Path
import argparse

# ==========================
# Configuration
# ==========================

parser = argparse.ArgumentParser(description="Clean dataset filenames and metadata.")
parser.add_argument("dir_path", type=str, help="Path to the source directory containing batches and metadata.csv")
args = parser.parse_args()

dir_path = args.dir_path

SOURCE_ROOT = Path(dir_path)          # folder containing batch_00x + metadata.csv
CSV_NAME = "metadata.csv"

DEST_ROOT = Path(dir_path+"_clean")
DEST_CSV_NAME = "metadata.csv"

REPLACEMENT_CHAR = "_"                 # character to replace special characters with

# ==========================
# Filename cleaning function
# ==========================

def clean_filename(filename: str, replacement: str = "_") -> str:
    """
    Replace special characters in filename while preserving extension.
    Allowed: letters, numbers, dot, underscore, dash
    """
    p = Path(filename)
    stem = p.stem
    suffix = p.suffix

    # Replace unwanted characters
    cleaned_stem = re.sub(r"[^A-Za-z0-9._]", replacement, stem)

    # Collapse multiple replacements
    cleaned_stem = re.sub(rf"{re.escape(replacement)}+", replacement, cleaned_stem)

    # Avoid empty names
    if not cleaned_stem:
        cleaned_stem = "file"

    return cleaned_stem + suffix


# ==========================
# Main process
# ==========================

def main():
    source_csv_path = SOURCE_ROOT / CSV_NAME
    dest_csv_path = DEST_ROOT / DEST_CSV_NAME

    DEST_ROOT.mkdir(parents=True, exist_ok=True)

    # Build mapping: original filename -> cleaned filename
    filename_map = {}

    rows = []

    with open(source_csv_path, newline="", encoding="utf-8") as f:
        reader = csv.reader(f)
        # header = next(reader)

        for row in reader:
            original_name = row[0]
            cleaned_name = clean_filename(original_name, REPLACEMENT_CHAR)

            filename_map[original_name] = cleaned_name

            new_row = row.copy()
            new_row[0] = original_name.split("/")[0] +"/" + cleaned_name
            # import pdb; pdb.set_trace()
            rows.append(new_row)

    # Copy and rename files batch by batch
    for batch_dir in SOURCE_ROOT.iterdir():
        if not batch_dir.is_dir():
            continue
        if not batch_dir.name.startswith("batch_"):
            continue

        dest_batch_dir = DEST_ROOT / batch_dir.name
        dest_batch_dir.mkdir(parents=True, exist_ok=True)

        for file_path in batch_dir.iterdir():
            if not file_path.is_file():
                continue

            original_name = file_path.name

            if str(Path(batch_dir.name) / original_name) not in filename_map:
                print(f"WARNING: {str(Path(batch_dir.name) / original_name)} not found in CSV. Skipping.")
                # import pdb; pdb.set_trace()
                continue

            cleaned_name = filename_map[str(Path(batch_dir.name) / original_name)]
            dest_file_path = dest_batch_dir / cleaned_name

            shutil.copy2(file_path, dest_file_path)

    # Write cleaned CSV
    with open(dest_csv_path, "w", newline="", encoding="utf-8") as f:
        writer = csv.writer(f)
        # writer.writerow(header)
        writer.writerows(rows)

    print("Clean dataset created successfully:")
    print(f"  Folder: {DEST_ROOT}")
    print(f"  CSV:    {dest_csv_path}")


if __name__ == "__main__":
    main()