scripts/coco_scene_paths.py

"""
Resolve COCO JSON paths under ``data/scenes/<scene>/`` after renames:

- ``_annotations_fixed.coco.json`` -> ``_annotations_original.coco.json``
- ``_annotations_supplementary.coco.json`` -> ``_annotations_extended.coco.json``

``extended`` adds annotations on the same images as ``original``; both may be present.
"""
from __future__ import annotations

from pathlib import Path
from typing import List, Optional

# Single canonical file for frame list / image dimensions (first match wins).
RESOLVE_ORDER: tuple[str, ...] = (
    "_annotations_original.coco.json",
    "_annotations_extended.coco.json",
    "_annotations_fixed.coco.json",
    "_annotations.coco.json",
)

# All COCO files to rasterize for per-instance masks (original then extended; skip legacy dup).
MASK_GENERATION_ORDER: tuple[str, ...] = (
    "_annotations_original.coco.json",
    "_annotations_extended.coco.json",
    "_annotations_fixed.coco.json",
)


def resolve_scene_coco_annotation(scene_dir: Path) -> Optional[Path]:
    """One annotation file for frame_id / H,W lookup."""
    for name in RESOLVE_ORDER:
        p = scene_dir / name
        if p.is_file():
            return p
    return None


def iter_scene_coco_for_masks(scene_dir: Path) -> List[Path]:
    """
    COCO files to generate instance_masks from. Uses ``_annotations.coco.json`` only if
    no renamed file exists (older scenes).
    """
    out: List[Path] = []
    for name in MASK_GENERATION_ORDER:
        p = scene_dir / name
        if p.is_file():
            out.append(p)
    if not out:
        p = scene_dir / "_annotations.coco.json"
        if p.is_file():
            out.append(p)
    return out


def open_coco_json(path: Path):
    import json

    for enc in ("utf-8-sig", "utf-8"):
        try:
            with path.open("r", encoding=enc) as f:
                return json.load(f)
        except UnicodeDecodeError:
            continue
    with path.open("r", encoding="utf-8", errors="replace") as f:
        return json.load(f)