loader.py

#!/usr/bin/env python3
"""Load VocalSet-Breath: pair the breath labels in this repo with your local
VocalSet audio.

VocalSet audio is not redistributed here. Download it from Zenodo
(DOI 10.5281/zenodo.1193957) and pass its path:

    python loader.py --vocalset /path/to/VocalSet

Or use it as a library:

    from loader import load_records
    records = load_records("labels", vocalset_dir="/path/to/VocalSet")
"""
from __future__ import annotations

import argparse
import json
from pathlib import Path


def load_records(labels_dir, vocalset_dir=None):
    """Return one record per labeled clip.

    Each record: clip_id, audio_path (if vocalset_dir given), duration_sec,
    breath_events as [(start_sec, end_sec), ...], and the full raw label dict.
    """
    labels_dir = Path(labels_dir)
    audio_index = _index_audio(Path(vocalset_dir)) if vocalset_dir else {}
    records = []
    for path in sorted(labels_dir.rglob("*.breath.json")):
        label = json.loads(path.read_text())
        clip_id = path.name.removesuffix(".breath.json")
        records.append({
            "clip_id": clip_id,
            "audio_path": audio_index.get(label.get("audio_file", clip_id + ".wav")),
            "duration_sec": label.get("duration_sec"),
            "breath_events": [(e["start_sec"], e["end_sec"])
                              for e in label.get("breath_events", [])],
            "label": label,
        })
    return records


def _index_audio(vocalset_dir):
    """Map filename -> absolute path for every .wav under a VocalSet tree."""
    return {p.name: str(p) for p in vocalset_dir.rglob("*.wav")}


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--labels", default="labels")
    ap.add_argument("--vocalset", default=None,
                    help="path to your local VocalSet download (audio is not shipped)")
    args = ap.parse_args()
    records = load_records(args.labels, args.vocalset)
    n_breaths = sum(len(r["breath_events"]) for r in records)
    print(f"{len(records)} clips · {n_breaths} breath events")
    if args.vocalset:
        missing = sum(1 for r in records if r["audio_path"] is None)
        print(f"{len(records) - missing} clips matched to audio, {missing} unmatched")


if __name__ == "__main__":
    main()