-
Notifications
You must be signed in to change notification settings - Fork 78
/
create_data.py
128 lines (118 loc) · 4.54 KB
/
create_data.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
import torch
from manopth import manolayer
import os
class DataSet:
def __init__(self, device=torch.device('cpu'), _mano_root='mano/models'):
args = {'flat_hand_mean': True, 'root_rot_mode': 'axisang',
'ncomps': 45, 'mano_root': _mano_root,
'no_pca': True, 'joint_rot_mode': 'axisang', 'side': 'right'}
self.mano = manolayer.ManoLayer(flat_hand_mean=args['flat_hand_mean'],
side=args['side'],
mano_root=args['mano_root'],
ncomps=args['ncomps'],
use_pca=not args['no_pca'],
root_rot_mode=args['root_rot_mode'],
joint_rot_mode=args['joint_rot_mode']
).to(device)
self.device = device
def new_cal_ref_bone(self, _shape):
parent_index = [0,
0, 1, 2,
0, 4, 5,
0, 7, 8,
0, 10, 11,
0, 13, 14
]
index = [0,
1, 2, 3, # index
4, 5, 6, # middle
7, 8, 9, # pinky
10, 11, 12, # ring
13, 14, 15] # thumb
reoder_index = [
13, 14, 15,
1, 2, 3,
4, 5, 6,
10, 11, 12,
7, 8, 9]
shape = _shape.clone().detach()
th_v_shaped = torch.matmul(self.mano.th_shapedirs,
shape.transpose(1, 0)).permute(2, 0, 1) \
+ self.mano.th_v_template
th_j = torch.matmul(self.mano.th_J_regressor, th_v_shaped)
temp1 = th_j.clone().detach()
temp2 = th_j.clone().detach()[:, parent_index, :]
result = temp1 - temp2
ref_len = th_j[:, [4], :] - th_j[:, [0], :]
ref_len = torch.norm(ref_len, dim=-1, keepdim=True)
result = torch.norm(result, dim=-1, keepdim=True)
result = result / ref_len
return torch.squeeze(result, dim=-1)[:, reoder_index]
def sample(self):
shape = 3 * torch.randn((1, 10))
result = self.new_cal_ref_bone(shape)
return (result, shape)
def batch_sample(self, batch_size):
shape = 3 * torch.randn((batch_size, 10))
result = self.new_cal_ref_bone(shape)
return (result, shape)
@staticmethod
def cal_ref_bone(_Jtr):
parent_index = [0,
0, 1, 2, 3,
0, 5, 6, 7,
0, 9, 10, 8,
0, 13, 14, 15,
0, 17, 18, 19
]
index = [1, 2, 3,
5, 6, 7,
9, 10, 11,
13, 14, 15,
17, 18, 19]
temp1 = _Jtr.clone().detach()
temp2 = _Jtr.clone().detach()[:, parent_index, :]
result = temp1 - temp2
result = result[:, index, :]
ref_len = _Jtr[:, [9], :] - _Jtr[:, [0], :]
ref_len = torch.norm(ref_len, dim=-1, keepdim=True)
result = torch.norm(result, dim=-1, keepdim=True)
# result = result / ref_len
return torch.squeeze(result, dim=-1)
if __name__ == '__main__':
dataset = DataSet()
import numpy as np
import tqdm
Total_Num = 1000000
NUM = 10000
data_bone = np.zeros((Total_Num, 15))
data_shape = np.zeros((Total_Num, 10))
for i in tqdm.tqdm(range(Total_Num // NUM)):
t1 = i * NUM
t2 = t1 + NUM
temp_1, temp_2 = dataset.batch_sample(NUM)
data_bone[t1:t2] = temp_1
data_shape[t1:t2] = temp_2
print(t1, t2)
save_dir = 'data'
if os.path.exists(save_dir):
pass
else:
os.mkdir(save_dir)
np.save(os.path.join(save_dir, 'data_bone.npy'), data_bone)
np.save(os.path.join(save_dir, 'data_shape.npy'), data_shape)
print('*' * 10, 'test', '*' * 10)
data_bone = np.load(os.path.join(save_dir, 'data_bone.npy'))
data_shape = np.load(os.path.join(save_dir, 'data_shape.npy'))
test_flag = 1
for i in tqdm.tqdm(range(Total_Num // NUM)):
t1 = i * NUM
t2 = t1 + NUM
test_shape = data_shape[t1:t2]
test_shape = torch.tensor(test_shape, dtype=torch.float)
test_bone = data_bone[t1:t2]
temp_1 = dataset.new_cal_ref_bone(test_shape)
flag = np.allclose(temp_1, test_bone)
flag = int(flag)
test_flag = test_flag * flag
print(test_flag)