-
Notifications
You must be signed in to change notification settings - Fork 9
/
train_pose_vae.py
63 lines (45 loc) · 1.7 KB
/
train_pose_vae.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
import time
import numpy as np
from glob import glob
from utils.vae import VAE, VAETrainer
from utils.feature import FeatureConverter, get_mocap_frames
if __name__ == '__main__':
from presets import preset
preset.load_default()
fc = FeatureConverter()
import argparse
parser = argparse.ArgumentParser(description='train pose vae ...')
parser.add_argument('--rawdata', type=str, default='data/vae/txt', help='raw txt motion data folder')
parser.add_argument('--epoch', type=int, default=80, help='num of epochs')
parser.add_argument('--name', type=str, default='test', help='name of the experiment')
args = parser.parse_args()
import time
stamp = time.strftime("%Y-%b-%d-%H%M%S", time.localtime())
exp_id = "%s-%s" % (args.name, stamp)
path = 'results/vae/' + exp_id
import os
if not os.path.exists(path):
os.makedirs(path)
data_file = '%s/data.npy' % path
import glob
motion_files = glob.iglob(args.rawdata + '/**/*.txt', recursive=True)
frames = []
for train_file in motion_files:
print(train_file)
cur_frames = get_mocap_frames(train_file)
for fr in cur_frames:
feature = fc.quat_to_pos(fr)
frames.append(feature)
frames.append(fc.mirror(feature))
data = np.array(frames)
np.random.shuffle(data)
with open(data_file, 'wb') as f:
np.save(f, data)
print('%d frames processed' % len(frames))
model = VAE(preset.vae.latent_dim)
trainer = VAETrainer(model, data)
if args.epoch is not None:
epochs = args.epoch
trainer.epochs = epochs
model_file = '%s/%s.pth' % (path, args.name)
trainer.train(model_file)