-
Notifications
You must be signed in to change notification settings - Fork 1
/
pose_estimation.py
175 lines (141 loc) · 7.12 KB
/
pose_estimation.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
import paddle
import paddle.nn as nn
import os
import sys
import math
import paddle.vision.models as models
import numpy as np
class Pose_Estimation(nn.Layer):
def __init__(self, net_dict, batch_norm=False):
super(Pose_Estimation, self).__init__()
self.model0 = self._make_layer(net_dict[0], batch_norm, True)
self.model1_1 = self._make_layer(net_dict[1][0], batch_norm)
self.model1_2 = self._make_layer(net_dict[1][1], batch_norm)
self.model2_1 = self._make_layer(net_dict[2][0], batch_norm)
self.model2_2 = self._make_layer(net_dict[2][1], batch_norm)
self.model3_1 = self._make_layer(net_dict[3][0], batch_norm)
self.model3_2 = self._make_layer(net_dict[3][1], batch_norm)
self.model4_1 = self._make_layer(net_dict[4][0], batch_norm)
self.model4_2 = self._make_layer(net_dict[4][1], batch_norm)
self.model5_1 = self._make_layer(net_dict[5][0], batch_norm)
self.model5_2 = self._make_layer(net_dict[5][1], batch_norm)
self.model6_1 = self._make_layer(net_dict[6][0], batch_norm)
self.model6_2 = self._make_layer(net_dict[6][1], batch_norm)
for m in self.sublayers():
if isinstance(m, nn.Conv2D):
n = m.weight.shape[0]*m.weight.shape[1]*m.weight.shape[2]
v = np.random.normal(loc=0.,scale=np.sqrt(2./n),size=m.weight.shape).astype('float32')
m.weight.set_value(v)
if m.bias is not None:
m.bias.set_value(np.zeros(m.bias.shape, dtype ='float32'))
elif isinstance(m, nn.BatchNorm2D):
m.weight.set_value(np.ones(m.weight.shape, dtype ='float32'))
m.bias.set_value(np.zeros(m.bias.shape, dtype ='float32'))
def _make_layer(self, net_dict, batch_norm=False, last_activity=False):
layers = []
length = len(net_dict) - 1
for i in range(length):
one_layer = net_dict[i]
key = list(one_layer.keys())[0]
v = one_layer[key]
if 'pool' in key:
layers += [nn.MaxPool2D(kernel_size=v[0], stride=v[1], padding=v[2])]
else:
conv2d = nn.Conv2D(in_channels=v[0], out_channels=v[1], kernel_size=v[2], stride=v[3], padding=v[4])
if batch_norm:
layers += [conv2d, nn.BatchNorm2D(v[1]), nn.ReLU()]
else:
layers += [conv2d, nn.ReLU()]
if last_activity:
one_layer = net_dict[-1]
key = list(one_layer.keys())[0]
v = one_layer[key]
conv2d = nn.Conv2D(in_channels=v[0], out_channels=v[1], kernel_size=v[2], stride=v[3], padding=v[4])
if batch_norm:
layers += [conv2d, nn.BatchNorm2D(v[1]), nn.ReLU()]
else:
layers += [conv2d, nn.ReLU()]
else:
one_layer = net_dict[-1]
key = list(one_layer.keys())[0]
v = one_layer[key]
conv2d = nn.Conv2D(in_channels=v[0], out_channels=v[1], kernel_size=v[2], stride=v[3], padding=v[4])
layers += [conv2d]
return nn.Sequential(*layers)
def forward(self, x):
out0 = self.model0(x)
out1_1 = self.model1_1(out0)
out1_2 = self.model1_2(out0)
out1 = paddle.concat([out1_1, out1_2, out0], 1)
out1_vec_mask = out1_1
out1_heat_mask = out1_2
out2_1 = self.model2_1(out1)
out2_2 = self.model2_2(out1)
out2 = paddle.concat([out2_1, out2_2, out0], 1)
out2_vec_mask = out2_1
out2_heat_mask = out2_2
out3_1 = self.model3_1(out2)
out3_2 = self.model3_2(out2)
out3 = paddle.concat([out3_1, out3_2, out0], 1)
out3_vec_mask = out3_1
out3_heat_mask = out3_2
out4_1 = self.model4_1(out3)
out4_2 = self.model4_2(out3)
out4 = paddle.concat([out4_1, out4_2, out0], 1)
out4_vec_mask = out4_1
out4_heat_mask = out4_2
out5_1 = self.model5_1(out4)
out5_2 = self.model5_2(out4)
out5 = paddle.concat([out5_1, out5_2, out0], 1)
out5_vec_mask = out5_1
out5_heat_mask = out5_2
out6_1 = self.model6_1(out5)
out6_2 = self.model6_2(out5)
out6_vec_mask = out6_1
out6_heat_mask = out6_2
return out1_vec_mask, out1_heat_mask, out2_vec_mask, out2_heat_mask, out3_vec_mask, out3_heat_mask, out4_vec_mask, out4_heat_mask, out5_vec_mask, out5_heat_mask, out6_vec_mask, out6_heat_mask
def PoseModel(num_point, num_vector, num_stages=6, batch_norm=False, pretrained=False):
net_dict = []
block0 = [{'conv1_1': [3, 64, 3, 1, 1]}, {'conv1_2': [64, 64, 3, 1, 1]}, {'pool1': [2, 2, 0]},
{'conv2_1': [64, 128, 3, 1, 1]}, {'conv2_2': [128, 128, 3, 1, 1]}, {'pool2': [2, 2, 0]},
{'conv3_1': [128, 256, 3, 1, 1]}, {'conv3_2': [256, 256, 3, 1, 1]}, {'conv3_3': [256, 256, 3, 1, 1]}, {'conv3_4': [256, 256, 3, 1, 1]}, {'pool3': [2, 2, 0]},
{'conv4_1': [256, 512, 3, 1, 1]}, {'conv4_2': [512, 512, 3, 1, 1]}, {'conv4_3_cpm': [512, 256, 3, 1, 1]}, {'conv4_4_cpm': [256, 128, 3, 1, 1]}]
net_dict.append(block0)
block1 = [[], []]
in_vec = [0, 128, 128, 128, 128, 512, num_vector * 2]
in_heat = [0, 128, 128, 128, 128, 512, num_point]
for i in range(1, 6):
if i < 4:
block1[0].append({'conv{}_stage1_vec'.format(i) :[in_vec[i], in_vec[i + 1], 3, 1, 1]})
block1[1].append({'conv{}_stage1_heat'.format(i):[in_heat[i], in_heat[i + 1], 3, 1, 1]})
else:
block1[0].append({'conv{}_stage1_vec'.format(i):[in_vec[i], in_vec[i + 1], 1, 1, 0]})
block1[1].append({'conv{}_stage1_heat'.format(i):[in_heat[i], in_heat[i + 1], 1, 1, 0]})
net_dict.append(block1)
in_vec_1 = [0, 128 + num_point + num_vector * 2, 128, 128, 128, 128, 128, 128, num_vector * 2]
in_heat_1 = [0, 128 + num_point + num_vector * 2, 128, 128, 128, 128, 128, 128, num_point]
for j in range(2, num_stages + 1):
blocks = [[], []]
for i in range(1, 8):
if i < 6:
blocks[0].append({'conv{}_stage{}_vec'.format(i, j):[in_vec_1[i], in_vec_1[i + 1], 7, 1, 3]})
blocks[1].append({'conv{}_stage{}_heat'.format(i, j):[in_heat_1[i], in_heat_1[i + 1], 7, 1, 3]})
else:
blocks[0].append({'conv{}_stage{}_vec'.format(i, j):[in_vec_1[i], in_vec_1[i + 1], 1, 1, 0]})
blocks[1].append({'conv{}_stage{}_heat'.format(i, j):[in_heat_1[i], in_heat_1[i + 1], 1, 1, 0]})
net_dict.append(blocks)
model = Pose_Estimation(net_dict, batch_norm)
if pretrained:
parameter_num = 10
vgg19_state_dict = paddle.load('vgg19.pdparams')
vgg19_keys = list(vgg19_state_dict.keys())
model_dict = model.state_dict()
from collections import OrderedDict
weights_load = OrderedDict()
for i in range(parameter_num):
weights_load[model.state_dict().keys()[i]] = vgg19_state_dict[vgg19_keys[i]]
model_dict.update(weights_load)
model.load_state_dict(model_dict)
return model
if __name__ == '__main__':
print(PoseModel(16, 14, 6, True, False))