-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathmodel.py
230 lines (195 loc) · 8.58 KB
/
model.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
import torch
import torch.nn as nn
import torch.nn.functional as F
import clip
class DoubleConv3D(nn.Module):
"""(convolution => [BN] => ReLU) * 2"""
def __init__(self, in_channels, out_channels, mid_channels=None):
super().__init__()
if not mid_channels:
mid_channels = out_channels
self.double_conv3d = nn.Sequential(
nn.Conv3d(in_channels, mid_channels, kernel_size=3, padding=1),
nn.BatchNorm3d(mid_channels),
nn.ReLU(inplace=True),
nn.Conv3d(mid_channels, out_channels, kernel_size=3, padding=1),
nn.BatchNorm3d(out_channels),
nn.ReLU(inplace=True)
)
def forward(self, x):
return self.double_conv3d(x)
class Down3D(nn.Module):
"""Downscaling with maxpool then double conv"""
def __init__(self, in_channels, out_channels):
super().__init__()
self.maxpool_conv = nn.Sequential(
nn.MaxPool3d(2),
DoubleConv3D(in_channels, out_channels)
)
def forward(self, x):
return self.maxpool_conv(x)
class Up3D(nn.Module):
"""Upscaling then double conv"""
def __init__(self, in_channels, out_channels, bilinear=True):
super().__init__()
# if bilinear, use the normal convolutions to reduce the number of channels
if bilinear:
self.up = nn.Upsample(scale_factor=2, mode='trilinear', align_corners=True)
self.conv = DoubleConv3D(in_channels, out_channels, in_channels // 2)
else:
self.up = nn.ConvTranspose3d(in_channels, in_channels // 2, kernel_size=2, stride=2)
self.conv = DoubleConv3D(in_channels, out_channels)
def forward(self, x1, x2):
x1 = self.up(x1)
# input is CHW
diffZ = x2.size()[2] - x1.size()[2]
diffY = x2.size()[3] - x1.size()[3]
diffX = x2.size()[4] - x1.size()[4]
x1 = F.pad(x1, [diffX // 2, diffX - diffX // 2,
diffY // 2, diffY - diffY // 2,
diffZ // 2, diffZ - diffZ // 2])
# if you have padding issues, see
# https://github.com/HaiyongJiang/U-Net-Pytorch-Unstructured-Buggy/commit/0e854509c2cea854e247a9c615f175f76fbb2e3a
# https://github.com/xiaopeng-liao/Pytorch-UNet/commit/8ebac70e633bac59fc22bb5195e513d5832fb3bd
x = torch.cat([x2, x1], dim=1)
return self.conv(x)
class OutConv3D(nn.Module):
def __init__(self, in_channels, out_channels):
super(OutConv3D, self).__init__()
self.conv = nn.Conv3d(in_channels, out_channels, kernel_size=1)
def forward(self, x):
return self.conv(x)
class UNet3D(nn.Module):
def __init__(self, inChannel, outChannel, fChannel=32, bilinear=True):
super(UNet3D, self).__init__()
self.inChannel = inChannel
self.outChannel = outChannel
self.fChannel=fChannel
self.bilinear = bilinear
self.inc = DoubleConv3D(inChannel, fChannel)
self.down1 = Down3D(fChannel, fChannel*2)
self.down2 = Down3D(fChannel*2, fChannel*4)
self.down3 = Down3D(fChannel*4, fChannel*8)
factor = 2 if bilinear else 1
self.down4 = Down3D(fChannel*8, fChannel*16 // factor)
self.up1 = Up3D(fChannel*16, fChannel*8 // factor, bilinear)
self.up2 = Up3D(fChannel*8, fChannel*4 // factor, bilinear)
self.up3 = Up3D(fChannel*4, fChannel*2 // factor, bilinear)
self.up4 = Up3D(fChannel*2, fChannel, bilinear)
self.outc = OutConv3D(fChannel, outChannel)
def forward(self, x,args=None):
x1 = self.inc(x)
x2 = self.down1(x1)
x3 = self.down2(x2)
x4 = self.down3(x3)
x5 = self.down4(x4)
x = self.up1(x5, x4)
x = self.up2(x, x3)
x = self.up3(x, x2)
x = self.up4(x, x1)
logits = self.outc(x)
return logits
def forward_x1(self,x):
x1 = self.inc(x)
return x1
class UNet3D_PSA(nn.Module):
def __init__(self, inChannel, outChannel, fChannel=32, bilinear=True):
super(UNet3D_PSA, self).__init__()
self.inChannel = inChannel
self.outChannel = outChannel
self.fChannel=fChannel
self.bilinear = bilinear
self.inc = DoubleConv3D(inChannel, fChannel)
self.down1 = Down3D(fChannel, fChannel*2)
self.down2 = Down3D(fChannel*2, fChannel*4)
self.down3 = Down3D(fChannel*4, fChannel*8)
factor = 2 if bilinear else 1
self.down4 = Down3D(fChannel*8, fChannel*16 // factor)
#decoder
self.up1 = nn.ModuleList([Up3D(fChannel*16, fChannel*8 // factor, bilinear) for i in range(5)])
self.up2 = nn.ModuleList([Up3D(fChannel*8, fChannel*4 // factor, bilinear)for i in range(5)])
self.up3 = nn.ModuleList([Up3D(fChannel*4, fChannel*2 // factor, bilinear)for i in range(5)])
self.up4 = nn.ModuleList([Up3D(fChannel*2, fChannel, bilinear)for i in range(5)])
self.outc = nn.ModuleList([OutConv3D(fChannel, outChannel)for i in range(5)])
self.train_decoder0=True
def forward(self, x, sentence):
if self.train_decoder0:
decoder=0
self.train_decoder0=False
else:
if sentence[-2]=='A':#Desai, Neil
decoder=1
elif sentence[-2]=='D':#Hannan, Raquibul
decoder=2
elif sentence[-2]=='I':#Yang, Daniel
decoder=3
elif sentence[-2]=='J':#Garant, Aurelie
decoder=4
else:
decoder=0 #Others
self.train_decoder0=True
x1 = self.inc(x)
x2 = self.down1(x1)
x3 = self.down2(x2)
x4 = self.down3(x3)
x5 = self.down4(x4)
x = self.up1[decoder](x5, x4)
x = self.up2[decoder](x, x3)
x = self.up3[decoder](x, x2)
x = self.up4[decoder](x, x1)
logits = self.outc[decoder](x)
return logits
class CLIPUNet3D(nn.Module):
def __init__(self, inChannel, outChannel, fChannel=32, bilinear=True):
super(CLIPUNet3D, self).__init__()
self.inChannel = inChannel
self.outChannel = outChannel
self.fChannel=fChannel
self.bilinear = bilinear
self.inc = DoubleConv3D(inChannel, fChannel)
self.down1 = Down3D(fChannel, fChannel*2)
self.down2 = Down3D(fChannel*2, fChannel*4)
self.down3 = Down3D(fChannel*4, fChannel*8)
factor = 2 if bilinear else 1
self.down4 = Down3D(fChannel*8, fChannel*16 // factor)
self.up1 = Up3D(fChannel*16, fChannel*8 // factor, bilinear)
self.up2 = Up3D(fChannel*8, fChannel*4 // factor, bilinear)
self.up3 = Up3D(fChannel*4, fChannel*2 // factor, bilinear)
self.up4 = Up3D(fChannel*2, fChannel, bilinear)
self.outc = OutConv3D(fChannel, outChannel)
self.clip_model, _ = clip.load("ViT-B/32", device='cpu')
self.downtext=nn.AvgPool1d(kernel_size=2,stride=2)
#text = clip.tokenize([r'There is no spacer hydrogel in the patient.',r'There is a spacer hydrogel in the patient.'])
#text = clip.tokenize([r'There is a type 2 spacer hydrogel in the patient.',r'There is no type 2 spacer hydrogel in the patient.'])
return
def forward(self, x, text):
x1 = self.inc(x)
x2 = self.down1(x1)
x3 = self.down2(x2)
x4 = self.down3(x3)
x5 = self.down4(x4)
text=clip.tokenize(text).to(x.device)
text_feature = self.clip_model.encode_text(text)
text_feature.unsqueeze_(dim=1).detach_()
text_feature=self.downtext(text_feature)
text_feature=(text_feature-0.015)/0.27
#text_feature=self.downtext(text_feature)
x5=x5*text_feature.view(1,text_feature.shape[2],1,1,1)
x4=x4*text_feature.view(1,text_feature.shape[2],1,1,1)
text_feature=self.downtext(text_feature)
x3=x3*text_feature.view(1,text_feature.shape[2],1,1,1)
text_feature=self.downtext(text_feature)
x2=x2*text_feature.view(1,text_feature.shape[2],1,1,1)
text_feature=self.downtext(text_feature)
x1=x1*text_feature.view(1,text_feature.shape[2],1,1,1)
x = self.up1(x5, x4)
x = self.up2(x, x3)
x = self.up3(x, x2)
x = self.up4(x, x1)
logits = self.outc(x)
return logits
if __name__ == '__main__':
model1=UNet3D_PSA(1,1,fChannel=32)
input=torch.randn(size=[1,1,64,64,64])
output1=model1(input,0)
pass