scripts/MS1M/process_ms1m.py

#!/usr/bin/env python3
"""Build MVEB MS-Celeb-1M subset from train/test split JSON + identity metadata."""

from __future__ import annotations

import argparse
import json
import shutil
import sys
from pathlib import Path
from typing import Dict, List, Tuple

_SCRIPTS_ROOT = Path(__file__).resolve().parent.parent
if str(_SCRIPTS_ROOT) not in sys.path:
    sys.path.insert(0, str(_SCRIPTS_ROOT))
from pack_media_parquet import pack_dataset_with_media, resolve_split_output_dir

QUERY_INSTRUCTION = "Represent the face with the following text."
QUERY_TEXT = "Retrieve all images with the same facial identity."
CANDIDATE_INSTRUCTION = "Represent the face with the following text."
CANDIDATE_TEXT = "Retrieve all images with the same facial identity."


def _load_path_to_instance_id(metadata_json: Path) -> Dict[str, str]:
    with metadata_json.open("r", encoding="utf-8") as f:
        annotations = json.load(f)
    if not isinstance(annotations, dict):
        raise ValueError(f"metadata must be a dict: {metadata_json}")

    path_to_inst: Dict[str, str] = {}
    for inst_id, image_names in annotations.items():
        if not isinstance(image_names, list):
            raise ValueError(f"metadata[{inst_id!r}] must be a list of image paths")
        inst = str(inst_id)
        for image_name in image_names:
            rel = str(image_name).replace("\\", "/")
            if rel in path_to_inst and path_to_inst[rel] != inst:
                raise ValueError(
                    f"Image path {rel!r} maps to multiple instance ids: "
                    f"{path_to_inst[rel]!r} and {inst!r}"
                )
            path_to_inst[rel] = inst
    return path_to_inst


def _make_annotations(
    rel_paths: List[str],
    path_to_inst: Dict[str, str],
) -> Tuple[List[dict], List[dict]]:
    candidate_rows: List[dict] = []
    inst_to_ids: Dict[str, List[str]] = {}
    missing: List[str] = []

    for idx, rel_path in enumerate(sorted(set(rel_paths))):
        rel = rel_path.replace("\\", "/")
        inst_id = path_to_inst.get(rel)
        if inst_id is None:
            missing.append(rel)
            continue
        candidate_rows.append(
            {
                "id": rel,
                "instance_id": inst_id,
                "image_path": rel,
                "instruction": CANDIDATE_INSTRUCTION,
                "text": CANDIDATE_TEXT,
            }
        )
        inst_to_ids.setdefault(inst_id, []).append(rel)

    if missing:
        preview = ", ".join(missing[:5])
        raise KeyError(
            f"{len(missing)} split image(s) not found in metadata.json "
            f"(e.g. {preview})"
        )

    query_rows: List[dict] = []
    for row in candidate_rows:
        target_ids = [x for x in inst_to_ids[row["instance_id"]] if x != row["id"]]
        if not target_ids:
            continue
        query_rows.append(
            {
                "id": row["id"],
                "instance_id": row["instance_id"],
                "image_path": row["image_path"],
                "instruction": QUERY_INSTRUCTION,
                "text": QUERY_TEXT,
                "target_ids": target_ids,
            }
        )
    return query_rows, candidate_rows


def _process_one_split(
    split_name: str,
    rel_paths: List[str],
    path_to_inst: Dict[str, str],
    image_root: Path,
    output_root: Path,
    overwrite: bool,
    media_rows_per_shard: int,
    row_group_size: int,
    num_workers: int,
) -> None:
    out_dir = resolve_split_output_dir(output_root, split_name, "MS-Celeb-1M")
    if overwrite and out_dir.exists():
        shutil.rmtree(out_dir)
    out_dir.mkdir(parents=True, exist_ok=True)

    query_rows, candidate_rows = _make_annotations(rel_paths, path_to_inst)
    stats = pack_dataset_with_media(
        query_annotations=query_rows,
        candidate_annotations=candidate_rows,
        image_dir=str(image_root),
        output_dir=str(out_dir),
        media_rows_per_shard=media_rows_per_shard,
        row_group_size=row_group_size,
        num_workers=num_workers,
        dataset_name="MS-Celeb-1M",
        data_split=split_name,
        write_subset_readme=True,
        show_progress=True,
    )
    print(
        f"[{split_name}] done: media={stats['num_media']}, "
        f"query={stats['num_query']}, candidate={stats['num_candidate']}, "
        f"shards={stats['num_shards']} -> {out_dir}"
    )


def main() -> None:
    script_dir = Path(__file__).resolve().parent
    root_dir = script_dir.parent.parent.parent

    parser = argparse.ArgumentParser(description="Process MS1M split JSON to MVEB parquet format.")
    parser.add_argument(
        "--split-json",
        type=Path,
        default=script_dir / "train_test_split.json",
        help="JSON containing train/test relative image paths.",
    )
    parser.add_argument(
        "--metadata-json",
        type=Path,
        default=script_dir / "metadata.json",
        help="Identity metadata: instance_id -> list of relative image paths.",
    )
    parser.add_argument(
        "--image-root",
        type=Path,
        default=root_dir / "source" / "ms1m",
        help="Root directory containing extracted shard folders 0000..0099.",
    )
    parser.add_argument(
        "--output-root",
        type=Path,
        default=root_dir,
        help="Output MVEB root directory (contains train/ and test/).",
    )
    parser.add_argument("--overwrite", action="store_true", help="Delete existing output split dir before writing.")
    parser.add_argument("--media-rows-per-shard", type=int, default=5000)
    parser.add_argument("--row-group-size", type=int, default=100)
    parser.add_argument("--num-workers", type=int, default=1)
    args = parser.parse_args()

    if not args.split_json.exists():
        raise FileNotFoundError(f"split json not found: {args.split_json}")
    if not args.metadata_json.exists():
        raise FileNotFoundError(f"metadata json not found: {args.metadata_json}")
    if not args.image_root.exists():
        raise FileNotFoundError(f"image root not found: {args.image_root}")

    print(f"==> Load metadata: {args.metadata_json}")
    path_to_inst = _load_path_to_instance_id(args.metadata_json)
    print(f"    images indexed: {len(path_to_inst)}")

    with args.split_json.open("r", encoding="utf-8") as f:
        split_data = json.load(f)
    for key in ("train", "test"):
        if key not in split_data or not isinstance(split_data[key], list):
            raise ValueError(f"split json must contain key {key!r} with a list value")

    _process_one_split(
        "train",
        split_data["train"],
        path_to_inst,
        args.image_root,
        args.output_root,
        args.overwrite,
        args.media_rows_per_shard,
        args.row_group_size,
        args.num_workers,
    )
    _process_one_split(
        "test",
        split_data["test"],
        path_to_inst,
        args.image_root,
        args.output_root,
        args.overwrite,
        args.media_rows_per_shard,
        args.row_group_size,
        args.num_workers,
    )


if __name__ == "__main__":
    main()