-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmcinet.py
209 lines (170 loc) · 6.66 KB
/
mcinet.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
# -*- coding: utf-8 -*-
#######################################################
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
from torch.nn.functional import softmax
from functools import partial
class EAHead(nn.Module): # External Attention
def __init__(self, c):# c为上一层的得到的通道数
super(EAHead, self).__init__()
# k初始值为32,我认为c表示的是最终将要输出的通道数
self.k = 128
self.first_conv = nn.Conv2d(c, c, 1)
self.k_linear = nn.Conv1d(c, self.k, 1, bias=False)
self.v_linear = nn.Conv1d(self.k, c, 1, bias=False)
def forward(self, x):
idn = x[:]
b, c, h, w = x.size()
x = self.first_conv(x)
x = x.view(b, c, -1) #
attn = self.k_linear(x)
attn = softmax(attn, dim=-1)
attn = attn / (attn.sum(dim=1, keepdim=True) + 1e-9)
x = self.v_linear(attn)
x = x.view(b, c, h, w)
x = x + idn
return x
class DoubleConv1(nn.Module):
def __init__(self, in_ch, out_ch,use_1x1conv=True):
super(DoubleConv1, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
# nn.BatchNorm2d(out_ch),
# nn.ReLU(inplace=True)
)
if use_1x1conv:
self.conv1 = nn.Conv2d(in_ch, out_ch, kernel_size=1, stride=1)
def forward(self, input):
Y = self.conv(input)
x = self.conv1(input)
return F.relu(Y+x)
class DoubleConv(nn.Module):
def __init__(self, in_ch, out_ch):
super(DoubleConv, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(in_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True),
nn.Conv2d(out_ch, out_ch, 3, padding=1),
nn.BatchNorm2d(out_ch),
nn.ReLU(inplace=True)
)
def forward(self, input):
x=self.conv(input)
return x
class BR(nn.Module):
def __init__(self, out_c):
super(BR, self).__init__()
# self.bn = nn.BatchNorm2d(out_c)
self.relu = nn.ReLU(inplace=True)
self.conv1 = nn.Conv2d(out_c, out_c, kernel_size=3, padding=1)
self.conv2 = nn.Conv2d(out_c, out_c, kernel_size=3, padding=1)
def forward(self, x):
# print('BR输入：',x.shape)
x_res = self.conv1(x)
x_res = self.relu(x_res)
x_res = self.conv2(x_res)
x = x + x_res
return x
# 空洞卷积模块
nonlinearity = partial(F.relu, inplace=True)
class DACblock(nn.Module):
def __init__(self, channel):
super(DACblock, self).__init__()
self.dilate2_2 = nn.Conv2d(channel, channel, kernel_size=3, dilation=2, padding=2)
self.dilate1 = nn.Conv2d(channel, channel, kernel_size=3, dilation=1, padding=1)
self.dilate2 = nn.Conv2d(channel, channel, kernel_size=3, dilation=3, padding=3)
self.dilate3 = nn.Conv2d(channel, channel, kernel_size=3, dilation=5, padding=5)
self.conv1x1 = nn.Conv2d(channel, channel, kernel_size=1, dilation=1, padding=0)
for m in self.modules():
if isinstance(m, nn.Conv2d) or isinstance(m, nn.ConvTranspose2d):
if m.bias is not None:
m.bias.data.zero_()
def forward(self, x):
# print('DAC输入：',x.shape)
dilate1_out = nonlinearity(self.dilate1(x))
dilate2_out = nonlinearity(self.conv1x1(self.dilate2(x)))
dilate3_out = nonlinearity(self.conv1x1(self.dilate2(self.dilate1(x))))
dilate4_out = nonlinearity(self.conv1x1(self.dilate3(self.dilate2_2(self.dilate1(x)))))
dilate5_out = nonlinearity(self.conv1x1(self.dilate2(self.dilate2_2(self.dilate1(x)))))
out = x + dilate1_out + dilate2_out + dilate3_out + dilate4_out+dilate5_out
return out
# 在每个跳跃连接层加入EA模块
class Mcinet(nn.Module):
def __init__(self, in_ch=3, out_ch=3):
super(Mcinet, self).__init__()
self.conv1 = DoubleConv1(in_ch, 32)
self.pool1 = nn.MaxPool2d(2)
self.conv2 = DoubleConv1(32, 64)
self.pool2 = nn.MaxPool2d(2)
self.conv3 = DoubleConv1(64, 128)
self.pool3 = nn.MaxPool2d(2)
self.conv4 = DoubleConv1(128, 256)
self.pool4 = nn.MaxPool2d(2)
self.conv5 = DoubleConv1(256, 512)
# center
self.dblock2 = DACblock(64)
self.dblock3 = DACblock(128)
self.dblock4 = DACblock(256)
self.dblock5 = DACblock(512)
# self.spp = SPPblock(512)
self.Att5 = EAHead(512)
self.br1=BR(64)
self.br2 = BR(128)
self.br3 = BR(256)
self.br4 = BR(512)
# decoder
self.conv6 = DoubleConv(512, 256)
self.conv7 = DoubleConv(256, 128)
self.conv8 = DoubleConv(128, 64)
self.conv9 = DoubleConv(64, 32)
self.up6 = nn.ConvTranspose2d(512, 256, 2, stride=2)
self.Att4 = EAHead(256)
self.up7 = nn.ConvTranspose2d(256, 128, 2, stride=2)
self.Att3 = EAHead(128)
self.up8 = nn.ConvTranspose2d(128, 64, 2, stride=2)
self.Att2 = EAHead(64)
self.up9 = nn.ConvTranspose2d(32, 3, 2, stride=2)
self.Att1 = EAHead(32)
self.conv10 = nn.Conv2d(3, out_ch, 1)
def forward(self, x):
c1 = self.conv1(x)
p1 = self.pool1(c1)
c2 = self.conv2(p1)
p2 = self.pool2(c2)
c3 = self.conv3(p2)
p3 = self.pool3(c3)
c4 = self.conv4(p3)
p4 = self.pool4(c4)
c5 = self.conv5(p4)
# center
c5 = self.dblock5(c5)
c5 = self.Att5(c5)
# c5 = self.spp(c5)
# decoding + concat path
up_6 = self.up6(c5)
A4 = self.Att4(self.dblock4(c4))
merge6 = torch.cat([A4, up_6], dim=1)
merge6=self.br4(merge6)
c6 = self.conv6(merge6)
up_7 = self.up7(c6)
A3 = self.Att3(self.dblock3(c3))
merge7 = torch.cat([A3, up_7], dim=1)
merge7 = self.br3(merge7)
c7 = self.conv7(merge7)
up_8 = self.up8(c7)
A2 = self.Att2(self.dblock2(c2))
merge8 = torch.cat([A2, up_8], dim=1)
merge8 = self.br2(merge8)
c8 = self.conv8(merge8)
up_9 = self.br1(c8)
up_9 = self.conv9(up_9)
up_9 = self.up9(up_9)
c10 = self.conv10(up_9)
out = nn.Sigmoid()(c10)
return out