scripts/Compcars/process_compcars.py

#!/usr/bin/env python3
"""Build MVEB CompCars subset from a precomputed train/test image split JSON.

Input JSON format:
{
  "train": ["1/1103/2011/xxx.jpg", ...],
  "test":  ["2/2201/2013/yyy.jpg", ...]
}

Output (when executed from MVEB root):
  ./MVEB-train/CompCars/{query.parquet,candidate.parquet,media-*.parquet,README.md}
  ./MVEB-test/CompCars/{query.parquet,candidate.parquet,media-*.parquet,README.md}
"""

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 image with the following text."
QUERY_TEXT = "Find a car of the same model, regardless of color."
CANDIDATE_INSTRUCTION = "You are a helpful assistant."
CANDIDATE_TEXT = "Find a car of the same model, regardless of color."


def _make_annotations(rel_paths: List[str]) -> Tuple[List[dict], List[dict]]:
    """Create query/candidate annotations compatible with pack_media_parquet."""
    candidate_rows: List[dict] = []
    inst_to_ids: Dict[str, List[str]] = {}

    for idx, rel_path in enumerate(rel_paths):
        parts = rel_path.split("/")
        if len(parts) < 4:
            raise ValueError(
                f"Invalid CompCars relative path: {rel_path!r}. "
                "Expected at least make/model/year/file.jpg"
            )
        inst_id = "_".join(parts[:3])
        row = {
            "id": rel_path,
            "instance_id": inst_id,
            "image_path": rel_path,
            "instruction": CANDIDATE_INSTRUCTION,
            "text": CANDIDATE_TEXT,
        }
        candidate_rows.append(row)
        inst_to_ids.setdefault(inst_id, []).append(rel_path)

    query_rows: List[dict] = []
    for row in candidate_rows:
        pos_ids = [x for x in inst_to_ids[row["instance_id"]] if x != row["id"]]
        if not pos_ids:
            # Skip single-image identities for query set, aligned with converter/comp_cars.py
            continue
        query_rows.append(
            {
                "id": row["id"],
                "instance_id": row["instance_id"],
                "image_path": row["image_path"],
                "instruction": QUERY_INSTRUCTION,
                "text": QUERY_TEXT,
                "pos_ids": pos_ids,
            }
        )

    return query_rows, candidate_rows


def _process_one_split(
    split_name: str,
    rel_paths: List[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, "CompCars")
    if overwrite and out_dir.exists():
        shutil.rmtree(out_dir)
    out_dir.mkdir(parents=True, exist_ok=True)

    unique_paths = sorted(set(rel_paths))
    query_rows, candidate_rows = _make_annotations(unique_paths)

    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="CompCars",
        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
    default_split_json = script_dir / "train_test_split.json"
    if not default_split_json.exists():
        # Backward-compatible filename used in some local copies.
        default_split_json = script_dir / "train_test_split.json"

    parser = argparse.ArgumentParser(description="Process CompCars split JSON to MVEB parquet format.")
    parser.add_argument(
        "--split-json",
        type=Path,
        default=default_split_json,
        help="JSON containing train/test relative image paths.",
    )
    parser.add_argument(
        "--image-root",
        type=Path,
        default=root_dir / "source" / "compcars" / "data" / "image",
        help="Root directory of raw CompCars images (contains make/model/year).",
    )
    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.image_root.exists():
        raise FileNotFoundError(f"image root not found: {args.image_root}")

    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"],
        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"],
        args.image_root,
        args.output_root,
        args.overwrite,
        args.media_rows_per_shard,
        args.row_group_size,
        args.num_workers,
    )


if __name__ == "__main__":
    main()