-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathconvert.py
230 lines (193 loc) · 11.5 KB
/
convert.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
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
import argparse
from model import Generator
from torch.autograd import Variable
import torch
import torch.nn.functional as F
import numpy as np
import os
from os.path import join, basename, dirname, split
import time
import datetime
from data_loader import to_categorical
import librosa
from utils import *
import glob
from tqdm import tqdm
# from data_loader import TestDataset, TestDataset2
emotions = ['sad', 'normal', 'angry']
emo2idx = dict(zip(emotions, range(len(emotions))))
target_speaker = 'liuchang'
def low_pass_filter(x, fs=int(1.0 / (5.0 * 0.001)), cutoff=20, padding=True):
"""FUNCTION TO APPLY LOW PASS FILTER
int(1.0 / (frame_period * 0.001))
Args:
x (ndarray): sequence
fs (int): Sampling frequency
cutoff (float): Cutoff frequency of low pass filter
Return:
(ndarray): Low pass filtered waveform sequence
"""
nyquist = fs // 2
norm_cutoff = cutoff / nyquist
# low cut filter
numtaps = 255
fil = firwin(numtaps, norm_cutoff)
x_pad = np.pad(x, (numtaps, numtaps), 'edge')
lpf_x = lfilter(fil, 1, x_pad)
lpf_x = lpf_x[numtaps + numtaps // 2: -numtaps // 2]
return lpf_x
class TestDataset(object):
"""Dataset for testing."""
def __init__(self, data_dir, src_wav_dir, trg_spk, src_emo, trg_emo):
# data_dir: */mc/test
# src_wav_dir: */liuchang_wavs_trimmed
self.trg_spk = trg_spk
self.trg_emo = trg_emo
self.src_emo = src_emo
self.src_spk_emo = '{}-{}'.format(trg_spk, src_emo) # e.g., liuchang-normal
self.trg_spk_emo = '{}-{}'.format(trg_spk, trg_emo) # e.g., liuchang-angry
self.mc_files = sorted(glob.glob(join(data_dir, '{}*.npy'.format(self.src_spk_emo))))
# load means and stds, which stored in the */mc/train dir
self.src_spk_stats = np.load(join(data_dir.replace('test', 'train'), '{}_stats.npz'.format(self.src_spk_emo)))
self.trg_spk_stats = np.load(join(data_dir.replace('test', 'train'), '{}_stats.npz'.format(self.trg_spk_emo)))
self.src_spk_stats_lf0_le = np.load(join(data_dir.replace('test', 'train').replace('mc', 'lf0_le'), '{}_stats.npz'.format(self.src_spk_emo)))
self.trg_spk_stats_lf0_le = np.load(join(data_dir.replace('test', 'train').replace('mc', 'lf0_le'), '{}_stats.npz'.format(self.trg_spk_emo)))
self.lf0_mean_src = self.src_spk_stats_lf0_le['log_f0s_mean']
self.lf0_std_src = self.src_spk_stats_lf0_le['log_f0s_std']
self.lf0_mean_trg = self.trg_spk_stats_lf0_le['log_f0s_mean']
self.lf0_std_trg = self.trg_spk_stats_lf0_le['log_f0s_std']
self.le_mean_src = self.src_spk_stats_lf0_le['log_energy_mean']
self.le_std_src = self.src_spk_stats_lf0_le['log_energy_std']
self.le_mean_trg = self.trg_spk_stats_lf0_le['log_energy_mean']
self.le_std_trg = self.trg_spk_stats_lf0_le['log_energy_std']
self.mcep_mean_src = self.src_spk_stats['coded_sps_mean']
self.mcep_std_src = self.src_spk_stats['coded_sps_std']
self.mcep_mean_trg = self.trg_spk_stats['coded_sps_mean']
self.mcep_std_trg = self.trg_spk_stats['coded_sps_std']
self.src_wav_dir = src_wav_dir
def get_batch_test_data(self, batch_size=8):
batch_data = []
for i in range(batch_size):
mcfile = self.mc_files[i]
filename = basename(mcfile).split('-')[-1]
wavfile_path = glob.glob(join(f"{self.src_wav_dir}/*/{self.trg_spk}/{self.src_emo}", filename.replace('npy', 'wav')))[0]
batch_data.append(wavfile_path)
return batch_data
def get_test_data(self):
data_list = []
for mcfile in self.mc_files:
filename = basename(mcfile).split('-')[-1]
wav_id = filename.split('.')[0]
if int(wav_id) < 200: continue
print(filename)
wavfile_path = glob.glob(join(f"{self.src_wav_dir}/{self.src_emo}", filename.replace('npy', 'wav')))[0]
data_list.append(wavfile_path)
return data_list
def load_wav(wavfile, sr=16000):
wav, _ = librosa.load(wavfile, sr=sr, mono=True)
return wav_padding(wav, sr=sr, frame_period=5, multiple = 4) # TODO
# return wav
def test(config):
current_dir = join(config.convert_dir, str(config.resume_iters))
os.makedirs(join(config.convert_dir, str(config.resume_iters)), exist_ok=True)
sampling_rate, num_mcep, frame_period=16000, 36, 5
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
G_A2B = Generator().to(device)
test_loader = TestDataset(config.test_data_dir,
config.src_wav_dir,
config.target_speaker,
config.source_emotion,
config.target_emotion)
# Restore model
print(f'Loading the trained models from step {config.resume_iters}...')
G_A2B_path = join(config.model_save_dir, f'{config.resume_iters}-G_A2B.ckpt')
# G_B2A_path = join(config.model_save_dir, f'{config.resume_iters}-G_B2A.ckpt')
G_A2B.load_state_dict(torch.load(G_A2B_path, map_location=lambda storage, loc: storage))
# G_B2A.load_state_dict(torch.load(G_B2A_path, map_location=lambda storage, loc: storage))
# Read a batch of testdata
# test_wavfiles = test_loader.get_batch_test_data(batch_size=8)
# test_wavs = [load_wav(wavfile, sampling_rate) for wavfile in test_wavfiles]
# Read testdata
test_wavfiles = test_loader.get_test_data()
# test_wavs = [load_wav(wavfile, sampling_rate) for wavfile in test_wavfiles]
print(f"Get {len(test_wavfiles)} test wav files!")
with torch.no_grad():
for wav_file_src in tqdm(test_wavfiles):
# print(len(wav))
wav_name = basename(wav_file_src) # 'id.wav'
wav_id = wav_name.split('.')[0]
# print(wav_name)
# ==== target wav ===== #
wav_file_trg = wav_file_src.replace(test_loader.src_emo, test_loader.trg_emo)
wav_trg = load_wav(wav_file_trg, sampling_rate)
f0_trg, _, sp_trg, _ = world_decompose(wav=wav_trg, fs=sampling_rate, frame_period=frame_period)
_, cont_lf0_lpf_trg = get_cont_lf0(f0_trg)
coded_sp_trg = world_encode_spectral_envelop(sp=sp_trg, fs=sampling_rate, dim=num_mcep)
trg_mc_filename = join(current_dir, f"{test_loader.trg_emo}-mc-{wav_id}.npy")
trg_lf0_filename = join(current_dir, f"{test_loader.trg_emo}-lf0-{wav_id}.npy")
np.save(file=trg_mc_filename, arr=coded_sp_trg, allow_pickle=False, fix_imports=True)
np.save(file=trg_lf0_filename, arr=cont_lf0_lpf_trg, allow_pickle=False, fix_imports=True)
# ===================== #
# ==== src wav and then convert ===== #
wav_src = load_wav(wav_file_src, sampling_rate)
f0, timeaxis, sp, ap = world_decompose(wav=wav_src, fs=sampling_rate, frame_period=frame_period)
# f0_converted = pitch_conversion(f0=f0,
# mean_log_src=test_loader.logf0s_mean_src, std_log_src=test_loader.logf0s_std_src,
# mean_log_target=test_loader.logf0s_mean_trg, std_log_target=test_loader.logf0s_std_trg)
uv, cont_lf0_lpf = get_cont_lf0(f0)
le = get_log_energy(sp)
lf0_normed = (cont_lf0_lpf - test_loader.lf0_mean_src) / test_loader.lf0_std_src
le_normed = (le - test_loader.le_mean_src) / test_loader.le_std_src
lf0_le_cwt = get_lf0_le_cwt(lf0_normed, le_normed)
coded_sp = world_encode_spectral_envelop(sp=sp, fs=sampling_rate, dim=num_mcep)
# print("Before being fed into G: ", coded_sp.shape)
coded_sp_norm = (coded_sp - test_loader.mcep_mean_src) / test_loader.mcep_std_src
mc_lf0_le = np.concatenate((coded_sp_norm, lf0_le_cwt), -1)
mc_lf0_le_tensor = torch.FloatTensor(mc_lf0_le.T).unsqueeze_(0).unsqueeze_(1).to(device)
# spk_emo_conds = torch.FloatTensor(test_loader.emo_c_trg).to(device)
# print(spk_emo_conds.size())
mc_lf0_le_converted_norm = G_A2B(mc_lf0_le_tensor).data.cpu().numpy()
mc_lf0_le_converted_norm = np.squeeze(mc_lf0_le_converted_norm).T
coded_sp_converted = mc_lf0_le_converted_norm[:, :36] * test_loader.mcep_std_trg + test_loader.mcep_mean_trg
coded_sp_converted = np.ascontiguousarray(coded_sp_converted)
lf0_converted = inverse_cwt(mc_lf0_le_converted_norm[:, 36:46].T) * test_loader.lf0_std_trg + test_loader.lf0_mean_trg
le_converted = inverse_cwt(mc_lf0_le_converted_norm[:, 46:].T) * test_loader.le_std_trg + test_loader.le_mean_trg
# ------------- low pass filter -------------
le_converted = low_pass_filter(le_converted)
# -------------------------------------------
e_converted = np.exp(le_converted)
f0_converted = np.squeeze(uv) * np.exp(lf0_converted)
decoded_sp_converted = world_decode_spectral_envelop(coded_sp_converted, sampling_rate)
e_sp_converted = np.linalg.norm(decoded_sp_converted, ord=2, axis=-1)
e_ratio = np.divide(e_converted, e_sp_converted)
# decoded_sp_converted = decoded_sp_converted * e_ratio[:, None]
wav_transformed = world_speech_synthesis(f0=f0_converted, sp=decoded_sp_converted,
ap=ap, fs=sampling_rate, frame_period=frame_period)
wav_id = wav_name.split('.')[0]
librosa.output.write_wav(join(current_dir,
f'{test_loader.src_spk_emo}-vcto-{test_loader.trg_emo}-{wav_id}.wav'), wav_transformed, sampling_rate)
converted_mc_filename = join(current_dir, f"converted-{test_loader.trg_emo}-mc-{wav_id}.npy")
converted_lf0_filename = join(current_dir, f"converted-{test_loader.trg_emo}-lf0-{wav_id}.npy")
np.save(file=converted_mc_filename, arr=coded_sp_converted, allow_pickle=False, fix_imports=True)
np.save(file=converted_lf0_filename, arr=lf0_converted, allow_pickle=False, fix_imports=True)
if [True, False][0]:
wav_cpsyn = world_speech_synthesis(f0=f0, sp=sp,
ap=ap, fs=sampling_rate, frame_period=frame_period)
librosa.output.write_wav(join(config.convert_dir, str(config.resume_iters), f'{test_loader.src_spk_emo}-cpsyn-{wav_name}'), wav_cpsyn, sampling_rate)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
# Model configuration.
parser.add_argument('--target_speaker', type=str, default='liuchang')
parser.add_argument('--source_emotion', type=str, default='normal')
parser.add_argument('--target_emotion', type=str, default='sad')
parser.add_argument('--resume_iters', type=int, default=18000, help='resume training from this step')
# Directories.
parser.add_argument('--train_data_dir', type=str, default='datasets/CASIA_dataset/mc/train')
parser.add_argument('--test_data_dir', type=str, default='datasets/CASIA_dataset/mc/test')
parser.add_argument('--src_wav_dir', type=str, default='datasets/CASIA_dataset/liuchang_wavs_trimmed')
parser.add_argument('--log_dir', type=str, default='./logs')
parser.add_argument('--model_save_dir', type=str, default='./logs/1026-0749-28-2018-mc-lf0-le-liuchang-normal-sad/models')
parser.add_argument('--convert_dir', type=str, default='./logs/1026-0749-28-2018-mc-lf0-le-liuchang-normal-sad/converted')
config = parser.parse_args()
print(config)
test(config)