-
Notifications
You must be signed in to change notification settings - Fork 14
/
Copy pathtest.py
201 lines (137 loc) · 6.28 KB
/
test.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
import pytorch_lightning as pl
import hydra
import torch
import os
import numpy as np
from lib.gdna_model import BaseModel
from tqdm import tqdm
import imageio
from lib.utils.render import render_mesh_dict, Renderer
import glob
from lib.dataset.datamodule import DataProcessor
from lib.model.helpers import Dict2Class
import pandas
@hydra.main(config_path="config", config_name="config")
def main(opt):
print(opt.pretty())
pl.seed_everything(opt.seed, workers=True)
torch.set_num_threads(10)
scan_info = pandas.read_csv(hydra.utils.to_absolute_path(opt.datamodule.data_list))
meta_info = Dict2Class({'n_samples': len(scan_info)})
data_processor = DataProcessor(opt.datamodule)
checkpoint_path = os.path.join('./checkpoints', 'last.ckpt')
model = BaseModel.load_from_checkpoint(
checkpoint_path=checkpoint_path,
strict=False,
opt=opt.model,
meta_info=meta_info,
data_processor=data_processor,
).cuda()
renderer = Renderer(256, anti_alias=True)
output_folder = 'videos'
if not os.path.exists(output_folder):
os.makedirs(output_folder)
max_samples = 200
smpl_param_zero = torch.zeros((1,86)).cuda().float()
smpl_param_zero[:,0] = 1
motion_folder = hydra.utils.to_absolute_path('data/aist_demo/seqs')
motion_files = sorted(glob.glob(os.path.join(motion_folder, '*.npz')))
smpl_param_anim = []
for f in motion_files:
f = np.load(f)
smpl_params = np.zeros(86)
smpl_params[0], smpl_params[4:76] = 1, f['pose']
smpl_param_anim.append(torch.tensor(smpl_params))
smpl_param_anim = torch.stack(smpl_param_anim).float().cuda()
batch_list = []
# prepare latent codes
if opt.eval_mode == 'z_shape':
idx_b = np.random.randint(0, meta_info.n_samples)
while len(batch_list) < max_samples:
idx_a = idx_b
idx_b = np.random.randint(0, meta_info.n_samples)
z_shape_a = model.z_shapes.weight.data[idx_a]
z_shape_b = model.z_shapes.weight.data[idx_b]
z_detail = model.z_details.weight.data.mean(0)
for i in range(10):
z_shape = torch.lerp(z_shape_a, z_shape_b, i/10)
batch = {'z_shape': z_shape[None],
'z_detail': z_detail[None],
'smpl_params': smpl_param_zero}
batch_list.append(batch)
if opt.eval_mode == 'z_detail':
idx_b = np.random.randint(0, meta_info.n_samples)
while len(batch_list) < max_samples:
idx_a = idx_b
idx_b = np.random.randint(0, meta_info.n_samples)
z_detail_a = model.z_details.weight.data[idx_a]
z_detail_b = model.z_details.weight.data[idx_b]
z_shape = model.z_shapes.weight.data.mean(0)
for i in range(10):
z_detail = torch.lerp(z_detail_a, z_detail_b, i/10)
batch = {'z_shape': z_shape[None],
'z_detail': z_detail[None],
'smpl_params': smpl_param_zero}
batch_list.append(batch)
if opt.eval_mode == 'betas':
z_shape = model.z_shapes.weight.data.mean(0)
z_detail = model.z_details.weight.data.mean(0)
betas = torch.cat([torch.linspace(0,-2,10),
torch.linspace(-2,0,10),
torch.linspace(0,2,10),
torch.linspace(2,0,10)])
for i in range(len(betas)):
smpl_param = smpl_param_zero.clone()
smpl_param[:, -10] = betas[i]
batch = {'z_shape': z_shape[None],
'z_detail': z_detail[None],
'smpl_params': smpl_param}
batch_list.append(batch)
if opt.eval_mode == 'thetas':
z_shape = model.z_shapes.weight.data.mean(0)
z_detail = model.z_details.weight.data.mean(0)
for i in range(len(smpl_param_anim)):
batch = {'z_shape': z_shape[None],
'z_detail': z_detail[None],
'smpl_params': smpl_param_anim[[i]]
}
batch_list.append(batch)
if opt.eval_mode == 'sample':
z_shapes, z_details = model.sample_codes(max_samples)
for i in range(len(z_shapes)):
id_smpl = np.random.randint(len(smpl_param_anim))
batch = {'z_shape': z_shapes[i][None],
'z_detail': z_details[i][None],
'smpl_params': smpl_param_anim[id_smpl][None],
}
batch_list.append(batch)
if opt.eval_mode == 'interp':
idx_b = np.random.randint(0, meta_info.n_samples)
while len(batch_list) < max_samples:
idx_a = idx_b
idx_b = np.random.randint(0, meta_info.n_samples)
z_shape_a = model.z_shapes.weight.data[idx_a]
z_shape_b = model.z_shapes.weight.data[idx_b]
z_detail_a = model.z_details.weight.data[idx_a]
z_detail_b = model.z_details.weight.data[idx_b]
for i in range(10):
z_shape = torch.lerp(z_shape_a, z_shape_b, i/10)
z_detail = torch.lerp(z_detail_a, z_detail_b, i/10)
batch = {'z_shape': z_shape[None],
'z_detail': z_detail[None],
'smpl_params': smpl_param_anim[len(batch_list)][None]}
batch_list.append(batch)
images_all = []
with torch.no_grad():
for i, batch in enumerate(tqdm(batch_list)):
cond = model.prepare_cond(batch)
batch_smpl = data_processor.process_smpl({'smpl_params': batch['smpl_params']}, model.smpl_server)
mesh_cano = model.extract_mesh(batch_smpl['smpl_verts_cano'], batch_smpl['smpl_tfs'], cond, res_up=4)
mesh_def = model.deform_mesh(mesh_cano, batch_smpl['smpl_tfs'])
img_def = render_mesh_dict(mesh_def,mode='xy',render_new=renderer)
img_def = np.concatenate([img_def[:256,:,:], img_def[256:,:,:]],axis=1)
images_all.append(img_def)
if i%10 == 0:
imageio.mimsave(os.path.join(output_folder,'%s_seed%d.mp4' % (opt.eval_mode,opt.seed)),images_all, codec='libx264')
if __name__ == '__main__':
main()