-
Notifications
You must be signed in to change notification settings - Fork 0
/
Res2NetModels.py
158 lines (136 loc) · 5.73 KB
/
Res2NetModels.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
import torch.nn as nn
import math
import torch.utils.model_zoo as model_zoo
import torch
import torch.nn.functional as F
__all__ = ['res2net50','res2net101','res2net152']
def conv3x3(in_planes, out_planes, stride=1, groups=1):
"""3x3 convolution with padding"""
return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
padding=1, groups=groups, bias=False)
def conv1x1(in_planes, out_planes, stride=1):
"""1x1 convolution"""
return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
class SEModule(nn.Module):
def __init__(self, channels, reduction=16):
super(SEModule, self).__init__()
self.avg_pool = nn.AdaptiveAvgPool2d(1)
self.fc1 = nn.Conv2d(channels, channels // reduction, kernel_size=1, padding=0)
self.relu = nn.ReLU(inplace=True)
self.fc2 = nn.Conv2d(channels // reduction, channels, kernel_size=1, padding=0)
self.sigmoid = nn.Sigmoid()
def forward(self, input):
x = self.avg_pool(input)
x = self.fc1(x)
x = self.relu(x)
x = self.fc2(x)
x = self.sigmoid(x)
return input * x
class Bottle2neck(nn.Module):
expansion = 4
def __init__(self, inplanes, outplanes, stride=1, downsample=None, scales = 4, se=None):
super(Bottle2neck, self).__init__()
self.conv1 = conv1x1(inplanes, outplanes, stride)
self.bn1 = nn.BatchNorm2d(outplanes)
self.conv2 = nn.ModuleList([conv3x3(outplanes//scales,outplanes//scales) for _ in range(scales - 1)])
self.bn2 = nn.ModuleList([nn.BatchNorm2d(outplanes//scales) for _ in range(scales - 1)])
self.conv3 = conv1x1(outplanes,outplanes*self.expansion)
self.bn3 = nn.BatchNorm2d(outplanes*self.expansion)
self.relu = nn.ReLU(inplace=True)
self.scales = scales
self.downsample = downsample
self.se = se
self.outplanes = outplanes
def forward(self,x):
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
xs = torch.split(out,self.outplanes//4,dim=1)
ys = []
for i in range(self.scales):
if i == 0:
ys.append(xs[i])
elif i==1:
ys.append(self.relu(self.bn2[i-1](self.conv2[i-1](xs[i]))))
else:
ys.append(self.relu(self.bn2[i-1](self.conv2[i-1](xs[i]+ys[-1]))))
out = torch.cat(ys,dim=1)
out = self.conv3(out)
out = self.bn3(out)
if self.se is not None:
out = self.se(out)
if self.downsample is not None:
x = self.downsample(x)
out = out+ x
out = self.relu(out)
return out
class Res2Net(nn.Module):
def __init__(self, num_block, num_class=200, width=16, scales=4, se=None):
super(Res2Net, self).__init__()
outplanes = [int(width*scales*2**i) for i in range(4)]
self.conv1 = nn.Conv2d(3, outplanes[0],kernel_size=7,stride=2,padding=3,bias=False)
self.bn1 = nn.BatchNorm2d(outplanes[0])
self.relu = nn.ReLU(inplace=True)
self.inplanes = outplanes[0]
self.maxpool = nn.MaxPool2d(kernel_size=3,stride=2,padding=1)
self.layer1 = self._make_layer(Bottle2neck, outplanes[0], num_block[0], scales=scales, se=se)
self.layer2 = self._make_layer(Bottle2neck, outplanes[1], num_block[1], stride=2, scales=scales, se=se)
self.layer3 = self._make_layer(Bottle2neck, outplanes[2], num_block[2], stride=2, scales=scales, se=se)
self.layer4 = self._make_layer(Bottle2neck, outplanes[3], num_block[3], stride=2, scales=scales, se=se)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.fc = nn.Linear(outplanes[3] * Bottle2neck.expansion, num_class)
self._initialize_weights()
def _make_layer(self,block,outplanes,num_block,stride=1,scales=4,se=None):
downsample = None
if stride != 1 or self.inplanes != outplanes * block.expansion:
downsample = nn.Sequential(
conv1x1(self.inplanes, outplanes * block.expansion, stride),
nn.BatchNorm2d(outplanes * block.expansion),
)
layers = []
layers.append(block(self.inplanes, outplanes, stride, downsample, scales=scales, se=se))
self.inplanes = outplanes * block.expansion
for _ in range(1,num_block):
layers.append(block(self.inplanes, outplanes, scales=scales, se=se))
return nn.Sequential(*layers)
def _initialize_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
if m.bias is not None:
m.bias.data.zero_()
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.Linear):
m.weight.data.normal_(0, 0.01)
m.bias.data.zero_()
def forward(self, x):
x = self.conv1(x)
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
x = self.avgpool(x)
x = x.view(x.size(0), -1)
x = self.fc(x)
return x
def res2net50(**kwargs):
"""Constructs a Res2Net-50 model.
"""
model = Res2Net([3, 4, 6, 3], **kwargs)
return model
def res2net101(**kwargs):
"""Constructs a ResNet-101 model.
"""
model = Res2Net([3, 4, 23, 3], **kwargs)
return model
def res2net152(**kwargs):
"""Constructs a ResNet-152 model.
"""
model = Res2Net([3, 8, 36, 3], **kwargs)
return model