forked from chengdazhi/Deformable-Convolution-V2-PyTorch
-
Notifications
You must be signed in to change notification settings - Fork 2
/
test_dcn3d_forward.py
180 lines (153 loc) · 6.31 KB
/
test_dcn3d_forward.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
#!/usr/bin/env python
from __future__ import absolute_import
from __future__ import print_function
from __future__ import division
import time
import torch
import torch.nn as nn
import numpy as np
from dcn.modules.deform_conv3d import DeformConv3d, _DeformConv3d, DeformConv3dPack
from dcn.deform_conv3d_naive import deform_conv3d_naive
deformable_groups = 1
N, inC, inD, inH, inW = 1, 4, 8, 16, 16
outC = 2
kD, kH, kW = 1, 3, 3
stride = 1
groups = 1
dilation = 1
padding = 1
torch.manual_seed(3)
def check_dconv_zero_offset():
conv_offset = nn.Conv3d(inC, deformable_groups * 3 * kD * kH * kW,
kernel_size=(kD, kH, kW),
stride=stride,
padding=padding,
dilation=dilation,
bias=True).cuda()
dcn = DeformConv3d(inC, outC, (kD, kH, kW),
stride=stride, padding=padding, dilation=dilation,
groups=groups,
deformable_groups=deformable_groups, im2col_step=1).cuda()
pcn = nn.Conv3d(inC, outC, (kD, kH, kW), stride=stride, padding=padding, dilation=dilation, groups=groups).cuda()
pcn.weight = dcn.weight
pcn.bias = dcn.bias
print((pcn.weight.data - dcn.weight.data).abs().max())
conv_offset.weight.data.zero_()
conv_offset.bias.data.zero_()
# conv_identify(dcn.weight, dcn.bias)
input = torch.randn(N, inC, inD, inH, inW).cuda()
offset = conv_offset(input)
output_d = dcn(input, offset)
output_p = pcn(input)
d = (output_d - output_p).abs().max()
if d < 1e-5:
print('dconv zero offset passed with {}'.format(d))
else:
print('dconv zero offset failed with {}'.format(d))
# print(output_p)
# print(output_d)
print((output_d - output_p).abs())
def check_dconv_naive_zero_offset():
conv_offset = nn.Conv3d(inC, deformable_groups * 3 * kD * kH * kW,
kernel_size=(kD, kH, kW),
stride=stride,
padding=padding,
dilation=dilation,
bias=True).cuda()
dcn = deform_conv3d_naive(inC, outC, (kD, kH, kW), stride=stride, padding=padding,
dilation=dilation, groups=groups, bias=False).cuda()
pcn = nn.Conv3d(inC, outC, (kD, kH, kW), stride=stride, padding=padding,
dilation=dilation, groups=groups, bias=False).cuda()
pcn.weight = dcn.weight
# pcn.bias = dcn.bias
print((pcn.weight.data - dcn.weight.data).abs().max())
conv_offset.weight.data.zero_()
conv_offset.bias.data.zero_()
# conv_identify(dcn.weight, dcn.bias)
input = torch.randn(N, inC, inD, inH, inW).cuda()
offset = conv_offset(input)
output_d = dcn(input, offset)
output_p = pcn(input)
d = (output_d - output_p).abs().max()
if d < 1e-5:
print('dconv zero offset passed with {}'.format(d))
else:
print('dconv zero offset failed with {}'.format(d))
# print(output_p)
# print(output_d)
print((output_d - output_p).abs())
def check_forward_dconv():
conv_offset = nn.Conv3d(inC, 1 * 3 * kD * kH * kW,
kernel_size=(kD, kH, kW),
stride=stride,
padding=padding,
dilation=dilation,
bias=True).cuda()
# conv_offset.weight.data.zero_()
# conv_offset.bias.data.zero_()
input = torch.randn(N, inC, inD, inH, inW).cuda()
offset = conv_offset(input)
dcn = DeformConv3d(inC, outC, (kD, kH, kW),
stride=stride, padding=padding, dilation=dilation,
groups=groups, bias=False).cuda()
dcnn = deform_conv3d_naive(inC, outC, (kD, kH, kW), stride=stride, padding=padding, dilation=dilation, groups=groups, bias=False).cuda()
dcnn.weight = dcn.weight
output1 = dcn(input, offset)
output2 = dcnn(input, offset)
d = (output1 - output2).abs().max()
if d < 1e-5:
print('dconv naive forward passed with {}'.format(d))
else:
print('dconv naive forward failed with {}'.format(d))
print(output1)
print(output2)
print((output1 - output2).abs())
def check_forward_dconv_mask():
conv_offset = nn.Conv3d(inC, 1 * 3 * kD * kH * kW,
kernel_size=(kD, kH, kW),
stride=stride,
padding=padding,
dilation=dilation,
bias=True).cuda()
# conv_offset.weight.data.zero_()
# conv_offset.bias.data.zero_()
input = torch.randn(N, inC, inD, inH, inW).cuda()
offset = conv_offset(input)
dim_mask = (0, 1, 1)
dim_mask = torch.Tensor(dim_mask).repeat(deformable_groups * kD * kH * kW)
dim_mask = dim_mask.view(1, 3 * deformable_groups * kD * kH * kW, 1, 1, 1).float()
offset = offset * dim_mask.cuda(input.get_device(), non_blocking=True)
offset = offset.view(N, deformable_groups, kD, kH, kW, 3, offset.size(2), offset.size(3), offset.size(4))
error = offset[:, :, :, :, :, 0, : ,: ,:].abs().sum()
if error == 0:
print('mask test passed')
else:
print('mask test failed with {}'.format(error))
if __name__ == '__main__':
check_dconv_naive_zero_offset()
check_forward_dconv()
check_forward_dconv_mask()
kernel_size_list = [1, 3, 5, 7]
stride_list = [1, 2]
padding_list = [0, 1, 2]
dilation_list = [1, 2]
for kernel_size in kernel_size_list:
print('kernel: {}'.format(kernel_size))
kH = kernel_size
kW = kernel_size
for stride_size in stride_list:
print('stride: {}'.format(stride_size))
stride = stride_size
for padding_size in padding_list:
print('padding: {}'.format(padding_size))
padding = padding_size
for dilation_size in dilation_list:
print('dilation: {}'.format(dilation_size))
dilation = dilation_size
check_dconv_naive_zero_offset()
check_forward_dconv()
# """
# ****** Note: backward is not reentrant error may not be a serious problem,
# ****** since the max error is less than 1e-7,
# ****** Still looking for what trigger this problem
# """