-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathattn_unet_2d.py
174 lines (139 loc) · 5.68 KB
/
attn_unet_2d.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
import copy
import torch
import torch.nn as nn
import pytorch_lightning as pl
class conv_block(pl.LightningModule):
def __init__(self, ch_in, ch_out):
super(conv_block, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(ch_in, ch_out, kernel_size=(3, 3), stride=(1, 1), padding=1),
nn.BatchNorm2d(ch_out),
nn.LeakyReLU(inplace=True),
nn.Conv2d(ch_out, ch_out, kernel_size=(3, 3), stride=(1, 1), padding=1),
nn.BatchNorm2d(ch_out),
nn.LeakyReLU(inplace=True)
)
def forward(self, x):
x = self.conv(x)
return x
class up_conv(pl.LightningModule):
def __init__(self, ch_in, ch_out):
super(up_conv, self).__init__()
self.up = nn.Sequential(
nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True),
nn.Conv2d(ch_in, ch_out, kernel_size=(3, 3), stride=(1, 1), padding=1),
nn.BatchNorm2d(ch_out),
nn.LeakyReLU(inplace=True)
)
def forward(self, x):
x = self.up(x)
return x
class Attention_block(pl.LightningModule):
def __init__(self, F_g, F_l, F_int):
super(Attention_block, self).__init__()
self.W_g = nn.Sequential(
nn.Conv2d(F_g, F_int, kernel_size=(1, 1), stride=(1, 1), padding=0),
nn.BatchNorm2d(F_int)
)
self.W_x = nn.Sequential(
nn.Conv2d(F_l, F_int, kernel_size=(1, 1), stride=(1, 1), padding=0),
nn.BatchNorm2d(F_int)
)
self.psi = nn.Sequential(
nn.Conv2d(F_int, 1, kernel_size=(1, 1), stride=(1, 1), padding=0),
nn.BatchNorm2d(1),
nn.Sigmoid()
)
self.relu = nn.ReLU(inplace=True)
def forward(self, g, x):
g1 = self.W_g(g)
x1 = self.W_x(x)
psi = self.relu(g1 + x1)
psi = self.psi(psi)
return x * psi
class AttU_Net2D(pl.LightningModule):
def __init__(self, drop):
super(AttU_Net2D, self).__init__()
self.Maxpool = nn.MaxPool2d(kernel_size=2, stride=2)
self.drop = nn.Dropout2d(p=drop)
self.upsample2d = nn.Upsample(scale_factor=(2, 2), mode='bilinear', align_corners=True)
self.Conv1 = conv_block(ch_in=1, ch_out=48)
self.Conv2 = conv_block(ch_in=48, ch_out=96)
self.Conv3 = conv_block(ch_in=96, ch_out=192)
self.Conv4 = conv_block(ch_in=192, ch_out=384)
self.Conv5 = conv_block(ch_in=384, ch_out=768)
self.Up5 = up_conv(ch_in=768, ch_out=384)
self.Att5 = Attention_block(F_g=384, F_l=384, F_int=192)
self.Up_conv5 = conv_block(ch_in=768, ch_out=384)
self.Up4 = up_conv(ch_in=384, ch_out=192)
self.Att4 = Attention_block(F_g=192, F_l=192, F_int=96)
self.Up_conv4 = conv_block(ch_in=384, ch_out=192)
self.Up3 = up_conv(ch_in=192, ch_out=96)
self.Att3 = Attention_block(F_g=96, F_l=96, F_int=48)
self.Up_conv3 = conv_block(ch_in=192, ch_out=96)
self.Up2 = up_conv(ch_in=96, ch_out=48)
self.Att2 = Attention_block(F_g=48, F_l=48, F_int=24)
self.Up_conv2 = conv_block(ch_in=96, ch_out=48)
self.Conv_1x1 = nn.Conv2d(48, 4, kernel_size=(1, 1))
# deep supervision 2nd decoder block
self.deep1 = nn.Conv2d(192, 4, kernel_size=(1, 1), padding='same')
# deep supervision 3rd decoder block
self.deep2 = nn.Conv2d(96, 4, kernel_size=(1, 1), padding='same')
self.neg_slope = 1e-2
self.apply(self.InitWeights_He)
def InitWeights_He(self, module):
if isinstance(module, nn.Conv3d) or isinstance(module, nn.Conv2d) or \
isinstance(module, nn.ConvTranspose2d) or isinstance(module, nn.ConvTranspose3d):
module.weight = nn.init.kaiming_normal_(module.weight, a=self.neg_slope)
if module.bias is not None:
module.bias = nn.init.constant_(module.bias, 0)
def forward(self, x):
# encoding path
x1 = self.Conv1(x)
x2 = self.Maxpool(x1)
x2 = self.Conv2(x2)
x2 = self.drop(x2) # dropout
x3 = self.Maxpool(x2)
x3 = self.Conv3(x3)
x3 = self.drop(x3) # dropout
x4 = self.Maxpool(x3)
x4 = self.Conv4(x4)
x4 = self.drop(x4) # dropout
x5 = self.Maxpool(x4)
x5 = self.Conv5(x5)
# decoding + concat path
d5 = self.Up5(x5)
x4 = self.Att5(g=d5, x=x4)
d5 = torch.cat((x4, d5), dim=1)
d5 = self.drop(d5)
d5 = self.Up_conv5(d5)
d4 = self.Up4(d5)
x3 = self.Att4(g=d4, x=x3)
d4 = torch.cat((x3, d4), dim=1)
d4 = self.drop(d4)
d4 = self.Up_conv4(d4)
ds2 = copy.copy(d4)
d3 = self.Up3(d4)
x2 = self.Att3(g=d3, x=x2)
d3 = torch.cat((x2, d3), dim=1)
d3 = self.drop(d3)
d3 = self.Up_conv3(d3)
ds3_2 = copy.copy(d3)
d2 = self.Up2(d3)
x1 = self.Att2(g=d2, x=x1)
d2 = torch.cat((x1, d2), dim=1)
d2 = self.Up_conv2(d2)
d1 = self.Conv_1x1(d2)
# Deep supervision
ds2_1x1_conv = self.deep1(ds2)
ds1_ds2_sum_upscale = self.upsample2d(ds2_1x1_conv)
ds3_1x1_conv = self.deep2(ds3_2)
ds1_ds2_sum_upscale_ds3_sum = torch.add(ds1_ds2_sum_upscale, ds3_1x1_conv)
ds1_ds2_sum_upscale_ds3_sum_upscale = self.upsample2d(ds1_ds2_sum_upscale_ds3_sum)
out = torch.add(d1, ds1_ds2_sum_upscale_ds3_sum_upscale)
return out
# if __name__ == "__main__":
# model = AttU_Net2D(0.0).cuda()
# inp = torch.rand(6, 1, 240, 240).cuda()
# output = model(inp)
# print("output", output.shape, "Number of parameters", sum(p.numel() for p in model.parameters()))