-
Notifications
You must be signed in to change notification settings - Fork 15
/
model_GateNet_ResNet.py
235 lines (198 loc) · 11.5 KB
/
model_GateNet_ResNet.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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
import torch
import torch.nn.functional as F
from torch import nn
from torch.utils import model_zoo
from torchvision.models.resnet import resnet50
from multi_scale_module import FoldConv_aspp
from thop import clever_format
from thop import profile
################################ResNet#######################################
__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
'resnet152']
model_urls = {
'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
}
def conv3x3(in_planes, out_planes, stride=1):
"""3x3 convolution with padding"""
return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
padding=1, bias=False)
class BasicBlock(nn.Module):
expansion = 1
def __init__(self, inplanes, planes, stride=1, downsample=None):
super(BasicBlock, self).__init__()
self.conv1 = conv3x3(inplanes, planes, stride)
self.bn1 = nn.BatchNorm2d(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv3x3(planes, planes)
self.bn2 = nn.BatchNorm2d(planes)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class Bottleneck(nn.Module):
expansion = 4
def __init__(self, inplanes, planes, stride=1, downsample=None,dilation=1):
super(Bottleneck, self).__init__()
self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1,bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,dilation=dilation,
padding=dilation, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)
self.bn3 = nn.BatchNorm2d(planes * 4)
self.relu = nn.ReLU(inplace=True)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.bn3(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
##########################################################################
class GateNet(nn.Module):
def __init__(self,block1,layers):
super(GateNet, self).__init__()
################################ResNet50---keep the last resolution#######################################
self.inplanes = 64
self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
bias=False)
self.bn1 = nn.BatchNorm2d(64)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block1, 64, layers[0])
self.layer2 = self._make_layer(block1, 128, layers[1], stride=2)
self.layer3 = self._make_layer(block1, 256, layers[2], stride=2)
#########################change the origional stride=2 to stride=1 aim to keep the resolution for fitting the FoldConv_aspp##############################################
self.layer4 = self._make_layer(block1, 512, layers[3], stride=1)
################################Gate#######################################
self.attention_feature5 = nn.Sequential(nn.Conv2d(64+64, 2, kernel_size=3, padding=1))
self.attention_feature4 = nn.Sequential(nn.Conv2d(256+64, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.PReLU(),
nn.Conv2d(64, 2, kernel_size=3, padding=1))
self.attention_feature3 = nn.Sequential(nn.Conv2d(512+128, 128, kernel_size=3, padding=1), nn.BatchNorm2d(128), nn.PReLU(),
nn.Conv2d(128, 2, kernel_size=3, padding=1))
self.attention_feature2 = nn.Sequential(nn.Conv2d(1024+256, 256, kernel_size=3, padding=1), nn.BatchNorm2d(256), nn.PReLU(),
nn.Conv2d(256, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.PReLU(),
nn.Conv2d(64, 2, kernel_size=3, padding=1))
self.attention_feature1 = nn.Sequential(nn.Conv2d(2048+512, 512, kernel_size=3, padding=1), nn.BatchNorm2d(512), nn.PReLU(),
nn.Conv2d(512, 128, kernel_size=3, padding=1), nn.BatchNorm2d(128), nn.PReLU(),
nn.Conv2d(128, 2, kernel_size=3, padding=1))
###############################Transition Layer########################################
self.dem1 = nn.Sequential(FoldConv_aspp(in_channel=2048,
out_channel=512,
out_size=384 // 16,
kernel_size=3,
stride=1,
padding=2,
dilation=2,
win_size=2,
win_padding=0,
), nn.BatchNorm2d(512), nn.PReLU())
self.dem2 = nn.Sequential(nn.Conv2d(1024, 256, kernel_size=3, padding=1), nn.BatchNorm2d(256), nn.PReLU())
self.dem3 = nn.Sequential(nn.Conv2d(512, 128, kernel_size=3, padding=1), nn.BatchNorm2d(128), nn.PReLU())
self.dem4 = nn.Sequential(nn.Conv2d(256, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.PReLU())
self.dem5 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.PReLU())
################################FPN branch#######################################
self.output1 = nn.Sequential(nn.Conv2d(512, 256, kernel_size=3, padding=1), nn.BatchNorm2d(256), nn.PReLU())
self.output2 = nn.Sequential(nn.Conv2d(256, 128, kernel_size=3, padding=1), nn.BatchNorm2d(128), nn.PReLU())
self.output3 = nn.Sequential(nn.Conv2d(128, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.PReLU())
self.output4 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=3, padding=1),nn.BatchNorm2d(64), nn.PReLU())
self.output5 = nn.Sequential(nn.Conv2d(64, 1, kernel_size=3, padding=1))
################################Parallel branch#######################################
self.out_res = nn.Sequential(nn.Conv2d(512+256+128+64+64+1, 256, kernel_size=3, padding=1), nn.BatchNorm2d(256), nn.PReLU(),
nn.Conv2d(256, 256, kernel_size=3, padding=1), nn.BatchNorm2d(256), nn.PReLU(),
nn.Conv2d(256, 1, kernel_size=3, padding=1))
#######################################################################
for m in self.modules():
if isinstance(m, nn.ReLU) or isinstance(m, nn.Dropout):
m.inplace = True
def forward(self, x):
input = x
B,_,_,_ = input.size()
################################Encoder block#######################################
x = self.conv1(x)
x = self.bn1(x)
E1 = self.relu(x)
x = self.maxpool(E1)
E2 = self.layer1(x)
E3 = self.layer2(E2)
E4 = self.layer3(E3)
E5 = self.layer4(E4)
################################Transition Layer#######################################
T5 = self.dem1(E5)
T4 = self.dem2(E4)
T3 = self.dem3(E3)
T2 = self.dem4(E2)
T1 = self.dem5(E1)
################################Gated FPN#######################################
G5 = self.attention_feature1(torch.cat((E5, T5), 1))
G5 = F.adaptive_avg_pool2d(F.sigmoid(G5), 1)
D5 = self.output1(G5[:, 0, :, :].unsqueeze(1).repeat(1, 512, 1, 1) * T5)
G4 = self.attention_feature2(torch.cat((E4,F.upsample(D5, size=E4.size()[2:], mode='bilinear')),1))
G4 = F.adaptive_avg_pool2d(F.sigmoid(G4),1)
D4 = self.output2(F.upsample(D5, size=E4.size()[2:], mode='bilinear')+G4[:, 0,:,:].unsqueeze(1).repeat(1,256,1,1)*T4)
G3 = self.attention_feature3(torch.cat((E3,F.upsample(D4, size=E3.size()[2:], mode='bilinear')),1))
G3 = F.adaptive_avg_pool2d(F.sigmoid(G3),1)
D3 = self.output3(F.upsample(D4, size=E3.size()[2:], mode='bilinear')+G3[:, 0,:,:].unsqueeze(1).repeat(1,128,1,1)*T3)
G2 = self.attention_feature4(torch.cat((E2,F.upsample(D3, size=E2.size()[2:], mode='bilinear')),1))
G2 = F.adaptive_avg_pool2d(F.sigmoid(G2),1)
D2 = self.output4(F.upsample(D3, size=E2.size()[2:], mode='bilinear')+G2[:, 0,:,:].unsqueeze(1).repeat(1,64,1,1)*T2)
G1 = self.attention_feature5(torch.cat((E1,F.upsample(D2, size=E1.size()[2:], mode='bilinear')),1))
G1 = F.adaptive_avg_pool2d(F.sigmoid(G1),1)
D1 = self.output5(F.upsample(D2, size=E1.size()[2:], mode='bilinear')+G1[:, 0,:,:].unsqueeze(1).repeat(1,64,1,1)*T1)
################################Gated Parallel&Dual branch residual fuse#######################################
output_fpn = F.upsample(D1, size=input.size()[2:], mode='bilinear')
output_res = self.out_res(torch.cat((D1,F.upsample(G5[:, 1,:,:].unsqueeze(1).repeat(1,512,1,1)*T5,size=E1.size()[2:], mode='bilinear'),F.upsample(G4[:, 1,:,:].unsqueeze(1).repeat(1,256,1,1)*T4,size=E1.size()[2:], mode='bilinear'),F.upsample(G3[:, 1,:,:].unsqueeze(1).repeat(1,128,1,1)*T3,size=E1.size()[2:], mode='bilinear'),F.upsample(G2[:, 1,:,:].unsqueeze(1).repeat(1,64,1,1)*T2,size=E1.size()[2:], mode='bilinear'),F.upsample(G1[:, 1,:,:].unsqueeze(1).repeat(1,64,1,1)*T1,size=E1.size()[2:], mode='bilinear')),1))
output_res = F.upsample(output_res,size=input.size()[2:], mode='bilinear')
pre_sal = output_fpn + output_res
#######################################################################
if self.training:
return output_fpn, pre_sal
# return F.sigmoid(pre_sal)
return pre_sal
def _make_layer(self, block, planes, blocks, stride=1,dilation=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=stride,bias=False),
nn.BatchNorm2d(planes * block.expansion),
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample,dilation))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
if __name__ == "__main__":
model = GateNet(Bottleneck,[3,4,6,3])
pretrained_dict = model_zoo.load_url(model_urls['resnet50'])
model_dict = model.state_dict()
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
model_dict.update(pretrained_dict)
model.load_state_dict(model_dict)
input = torch.autograd.Variable(torch.randn(4, 3, 384, 384))
output = model(input)