File size: 1,685 Bytes
9ade843
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
"""Add mic-array DOA to every non-speech sound event of every protocol clip (needs official VRS)."""
import json, os, sys, numpy as np
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
from doa import ArrayDOA
H = os.path.dirname(os.path.abspath(__file__))
B = '/mnt/work_new/weihanxu/aea_filtered_1105'
sc = json.load(open(f'{H}/scores.json'))
byrec = {}
for c in sc: byrec.setdefault(c['recording'], []).append(c)
for rec, clips in byrec.items():
    vrs = f'/mnt/data_new/AEA/{rec}/recording.vrs'
    if not os.path.exists(vrs): print('NO VRS', rec); continue
    tg = json.load(open(f'{B}/annotations/legacy/traj_gaze/{rec}.json'))
    t0 = float(tg[0]['tracking_timestamp']) - tg[0]['frame_index'] / 20.0
    D = ArrayDOA(vrs)
    for c in clips:
        for e in c['sound_events']:
            s, t = e['start'] + t0, e['end'] + t0
            if t - s < 0.4: s, t = s - 0.2, s + 0.4   # short impulsive sounds: onset window
            t = min(t, s + 3.0)
            r = D.estimate(s, t)
            e['doa'] = r and {k: round(v, 2) for k, v in r.items()}
        ok = [e for e in c['sound_events'] if e.get('doa') and e['doa']['peak_ratio'] >= 1.5]
        off = [e for e in ok if abs(e['doa']['az_deg']) > 45]
        c['n_doa_events'] = len(ok); c['n_offaxis_doa'] = len(off)
        c['frac_offaxis_doa'] = round(len(off) / len(ok), 2) if ok else 0.0
        c['offaxis_labels'] = sorted({e['label'] for e in off})
        print(f"{c['seg_id']:34s} doa={len(ok):3d} off={len(off):3d} frac={c['frac_offaxis_doa']:.2f} {c['offaxis_labels'][:5]}", flush=True)
json.dump(sc, open(f'{H}/scores_doa.json', 'w'), indent=1, ensure_ascii=False)
print('SCORE_DOA_DONE')