-
Notifications
You must be signed in to change notification settings - Fork 1
/
evaluate.py
executable file
·77 lines (58 loc) · 4.28 KB
/
evaluate.py
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
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
import os
import librosa
import argparse
import statistics as stat
import numpy as np
from metrics_binaural import compute_metrics
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--num_mix', type=int, required=True)
args = parser.parse_args()
spec_stft_distance_list = []
spec_envelope_distance_list = []
wave_stft_distance_list = []
wave_envelope_distance_list = []
mono_stft_distance_list = []
mono_envelope_distance_list = []
shift_stft_distance_list = []
shift_envelope_distance_list = []
num_mix = args.num_mix
test_data_path = os.path.join(os.path.split(os.path.join(os.getcwd(), __file__))[0], 'small_test_data',
'Source_N_'+str(num_mix))
folders = os.listdir(test_data_path)
for folder in folders:
#define paths
predicted_binaural_spec_path = os.path.join(test_data_path, folder, 'pred_Mono2Binaural_rgbs.wav')
predicted_binaural_wave_path = os.path.join(test_data_path, folder, 'pred_Points2Sound_rgbs.wav')
gt_binaural_path = os.path.join(test_data_path, folder, 'gt.wav')
shift_path = os.path.join(test_data_path, folder, 'pred_rotate_Points2Sound_rgbs.wav')
print(gt_binaural_path)
#load audios
spec_predicted_binaural, _ = librosa.load(predicted_binaural_spec_path, sr=44100, mono=False)
wave_predicted_binaural, _ = librosa.load(predicted_binaural_wave_path, sr=44100, mono=False)
gt_audio_binaural, _ = librosa.load(gt_binaural_path, sr=44100, mono=False)
mono_audio = gt_audio_binaural[0, :] + gt_audio_binaural[1, :]
mono_audio = np.repeat(np.expand_dims(mono_audio, 0), 2, axis=0)
shift_audio, _ = librosa.load(shift_path, sr=44100, mono=False)
stft_l2_spec, envelope_distance_spec = compute_metrics(spec_predicted_binaural, gt_audio_binaural)
spec_stft_distance_list.append(stft_l2_spec)
spec_envelope_distance_list.append(envelope_distance_spec)
stft_l2_wave, envelope_distance_wave = compute_metrics(wave_predicted_binaural, gt_audio_binaural)
wave_stft_distance_list.append(stft_l2_wave)
wave_envelope_distance_list.append(envelope_distance_wave)
stft_l2_mono, envelope_distance_mono = compute_metrics(mono_audio, gt_audio_binaural)
mono_stft_distance_list.append(stft_l2_mono)
mono_envelope_distance_list.append(envelope_distance_mono)
stft_l2_shift, envelope_distance_shift = compute_metrics(shift_audio, gt_audio_binaural)
shift_stft_distance_list.append(stft_l2_shift)
shift_envelope_distance_list.append(envelope_distance_shift)
print("MONO2BINAURAL STFT L2 Distance: ", stat.mean(spec_stft_distance_list), stat.stdev(spec_stft_distance_list), stat.stdev(spec_stft_distance_list) / np.sqrt(len(spec_stft_distance_list)))
print("MONO2BINAURAL Average Envelope Distance: ", stat.mean(spec_envelope_distance_list), stat.stdev(spec_envelope_distance_list), stat.stdev(spec_envelope_distance_list) / np.sqrt(len(spec_envelope_distance_list)))
print("POINTS2SOUND STFT L2 Distance: ", stat.mean(wave_stft_distance_list), stat.stdev(wave_stft_distance_list), stat.stdev(wave_stft_distance_list) / np.sqrt(len(wave_stft_distance_list)))
print("POINTS2SOUND Average Envelope Distance: ", stat.mean(wave_envelope_distance_list), stat.stdev(wave_envelope_distance_list), stat.stdev(wave_envelope_distance_list) / np.sqrt(len(wave_envelope_distance_list)))
print("MONO-MONO STFT L2 Distance: ", stat.mean(mono_stft_distance_list), stat.stdev(mono_stft_distance_list), stat.stdev(mono_stft_distance_list) / np.sqrt(len(mono_stft_distance_list)))
print("MONO-MONO Average Envelope Distance: ", stat.mean(mono_envelope_distance_list), stat.stdev(mono_envelope_distance_list), stat.stdev(mono_envelope_distance_list) / np.sqrt(len(mono_envelope_distance_list)))
print("ROTATED STFT L2 Distance: ", stat.mean(shift_stft_distance_list), stat.stdev(shift_stft_distance_list), stat.stdev(shift_stft_distance_list) / np.sqrt(len(shift_stft_distance_list)))
print("ROTATED Average Envelope Distance: ", stat.mean(shift_envelope_distance_list), stat.stdev(shift_envelope_distance_list), stat.stdev(shift_envelope_distance_list) / np.sqrt(len(shift_envelope_distance_list)))
if __name__ == '__main__':
main()