-
Notifications
You must be signed in to change notification settings - Fork 1
/
torchvision_backbones.py
176 lines (139 loc) · 5.64 KB
/
torchvision_backbones.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
"""
Backbones supported by torchvison.
"""
from collections import OrderedDict
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
import torch
import torch.nn as nn
import torch.nn.functional as F
class BasicBlock(nn.Module):
def __init__(self, in_planes, planes, stride=1):
super(BasicBlock, self).__init__()
self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.shortcut = nn.Sequential()
if stride != 1 or in_planes != planes:
self.shortcut = nn.Sequential(
nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(planes)
)
# SE layers
self.fc1 = nn.Conv2d(planes, planes//16, kernel_size=1) # Use nn.Conv2d instead of nn.Linear
self.fc2 = nn.Conv2d(planes//16, planes, kernel_size=1)
def forward(self, x):
out = F.relu(self.bn1(self.conv1(x)))
out = self.bn2(self.conv2(out))
# Squeeze
w = F.avg_pool2d(out, out.size(2))
w = F.relu(self.fc1(w))
w = F.sigmoid(self.fc2(w))
# Excitation
out = out * w # New broadcasting feature from v0.2!
out += self.shortcut(x)
out = F.relu(out)
return out
class PreActBlock(nn.Module):
def __init__(self, in_planes, planes, stride=1):
super(PreActBlock, self).__init__()
self.bn1 = nn.BatchNorm2d(in_planes)
self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
if stride != 1 or in_planes != planes:
self.shortcut = nn.Sequential(
nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride, bias=False)
)
# SE layers
self.fc1 = nn.Conv2d(planes, planes//16, kernel_size=1)
self.fc2 = nn.Conv2d(planes//16, planes, kernel_size=1)
def forward(self, x):
out = F.relu(self.bn1(x))
shortcut = self.shortcut(out) if hasattr(self, 'shortcut') else x
out = self.conv1(out)
out = self.conv2(F.relu(self.bn2(out)))
# Squeeze
w = F.avg_pool2d(out, out.size(2))
w = F.relu(self.fc1(w))
w = F.sigmoid(self.fc2(w))
# Excitation
out = out * w
out += shortcut
return out
class SENet(nn.Module):
def __init__(self, block, num_blocks, num_classes=10):
super(SENet, self).__init__()
self.in_planes = 64
self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)
self.bn1 = nn.BatchNorm2d(64)
self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1)
self.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)
self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)
self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)
self.linear = nn.Linear(512, num_classes)
def _make_layer(self, block, planes, num_blocks, stride):
strides = [stride] + [1]*(num_blocks-1)
layers = []
for stride in strides:
layers.append(block(self.in_planes, planes, stride))
self.in_planes = planes
return nn.Sequential(*layers)
def forward(self, x):
out = F.relu(self.bn1(self.conv1(x)))
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
out = self.layer4(out)
out = F.avg_pool2d(out, 4)
out = out.view(out.size(0), -1)
out = self.linear(out)
return out
def SENet18():
return SENet(PreActBlock, [2,2,2,2])
# def test():
# net = SENet18()
# y = net(torch.randn(1,3,32,32))
# print(y.size())
# test()
class TVDeeplabRes101Encoder(nn.Module):
"""
FCN-Resnet101 backbone from torchvision deeplabv3
No ASPP is used as we found emperically it hurts performance
"""
def __init__(self, use_coco_init, aux_dim_keep = 64, use_aspp = False):
super().__init__()
_model = torchvision.models.segmentation.deeplabv3_resnet101(pretrained=use_coco_init, progress=True, num_classes=21, aux_loss=None)
if use_coco_init:
print("###### NETWORK: Using ms-coco initialization ######")
else:
print("###### NETWORK: Training from scratch ######")
_model_list = list(_model.children())
self.aux_dim_keep = aux_dim_keep
self.backbone = _model_list[0]
self.localconv = nn.Conv2d(2048, 256,kernel_size = 1, stride = 1, bias = False) # reduce feature map dimension
self.asppconv = nn.Conv2d(256, 256,kernel_size = 1, bias = False)
_aspp = _model_list[1][0]
_conv256 = _model_list[1][1]
self.aspp_out = nn.Sequential(*[_aspp, _conv256] )
self.use_aspp = use_aspp
def forward(self, x_in, low_level):
"""
Args:
low_level: whether returning aggregated low-level features in FCN
"""
fts = self.backbone(x_in)
if self.use_aspp:
fts256 = self.aspp_out(fts['out'])
high_level_fts = fts256
else:
fts2048 = fts['out']
high_level_fts = self.localconv(fts2048)
if low_level:
low_level_fts = fts['aux'][:, : self.aux_dim_keep]
return high_level_fts, low_level_fts
else:
return high_level_fts