-
Notifications
You must be signed in to change notification settings - Fork 13
/
prepare_meta_from_tfr.py
64 lines (55 loc) · 2.21 KB
/
prepare_meta_from_tfr.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
import os
import numpy as np
import tensorflow as tf
from tqdm import tqdm
from text import sequence_to_text
from tfr_dset import load_for_prepare_meta_from_tfr
max_steps = 100000 # actual 95650
tfr_dir = 'bc2013/training/tfrs_with_emo_feature'
meta_path = 'bc2013/full_meta.txt'
mel_path = 'bc2013/mels'
spec_path = 'bc2013/specs'
def main():
tf_dset = load_for_prepare_meta_from_tfr(tfr_dir)
feats = tf_dset.make_one_shot_iterator().get_next()
i, lines, mels, specs = 1, [], [], []
lines.append('# emo|aro|val|text_len|spec_len|text|uid|mel|spec\n')
pbar = tqdm(max_steps)
sess = tf.Session()
try:
while True:
fetched_feats = sess.run(feats)
# uid = fetched_feats['uid'].tobytes().decode('utf-8')
uid = fetched_feats['uid'].decode('utf-8')
text = sequence_to_text(fetched_feats['inputs'])
text_lens = fetched_feats['input_lengths']
mel = fetched_feats['mel_targets']
spec = fetched_feats['linear_targets']
spec_lens = fetched_feats['spec_lengths']
emo = '[{:.5f}, {:.5f}, {:.5f}, {:.5f}]'.format(*fetched_feats['soft_emo_labels'])
aro = '[{:.5f}, {:.5f}]'.format(*fetched_feats['soft_arousal_labels'])
val = '[{:.5f}, {:.5f}]'.format(*fetched_feats['soft_valance_labels'])
mel_name = os.path.join(mel_path, f'bc13-mel-{i:06d}.npy')
spec_name = os.path.join(spec_path, f'bc13-spec-{i:06d}.npy')
line = f'{emo}|{aro}|{val}|{text_lens}|{spec_lens}|{text}|{uid}|{mel_name}|{spec_name}\n'
lines.append(line)
mels.append([mel, mel_name])
specs.append([spec, spec_name])
pbar.update(1)
i += 1
except tf.errors.OutOfRangeError:
print('sess.run finished!')
finally:
pbar.close()
sess.close()
os.makedirs(mel_path, exist_ok=True)
os.makedirs(spec_path, exist_ok=True)
for mel, mel_name in tqdm(mels):
np.save(mel_name, mel)
for spec, spec_name in tqdm(specs):
np.save(spec_name, spec)
with open(meta_path, 'w') as fw:
fw.writelines(lines)
print(f'total {i} items finished!')
if __name__ == '__main__':
main()