SACuP/model.py at main · AIRLABkhu/SACuP · GitHub

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
# full assembly of the sub-parts to form the complete net

import torch.nn.functional as F

from model_parts import *
from torchvision import models

class UNet(nn.Module):
    def __init__(self, in_channels, n_classes):
        super(UNet, self).__init__()
        self.inc = inconv(in_channels, 64)
        self.down1 = down(64, 128)
        self.down2 = down(128, 256)
        self.down3 = down(256, 512)
        self.down4 = down(512, 512)
        self.up1 = up(1024, 256)
        self.up2 = up(512, 128)
        self.up3 = up(256, 64)
        self.up4 = up(128, 64)
        self.outc = outconv(64, n_classes)

    def forward(self, x):
        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)
        x = self.outc(x)
        return x

def up_conv(in_channels, out_channels):
    return nn.ConvTranspose2d(in_channels, out_channels, kernel_size=2, stride=2)

class ResNetUNet(nn.Module):
    """Shallow Unet with ResNet18 or ResNet34 encoder."""

    def __init__(self, in_channels, n_classes=12, encoder=models.resnet34):
        super().__init__()
        self.encoder = encoder(pretrained=False)
        self.encoder.conv1 = nn.Conv2d(in_channels, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
        self.encoder_layers = list(self.encoder.children())

        self.block1 = nn.Sequential(*self.encoder_layers[:3])
        self.block2 = nn.Sequential(*self.encoder_layers[3:5])
        self.block3 = self.encoder_layers[5]
        self.block4 = self.encoder_layers[6]
        self.block5 = self.encoder_layers[7]

        self.up_conv6 = up_conv(512, 512)
        self.conv6 = double_conv(512 + 256, 512)
        self.up_conv7 = up_conv(512, 256)
        self.conv7 = double_conv(256 + 128, 256)
        self.up_conv8 = up_conv(256, 128)
        self.conv8 = double_conv(128 + 64, 128)
        self.up_conv9 = up_conv(128, 64)
        self.conv9 = double_conv(64 + 64, 64)
        self.up_conv10 = up_conv(64, 32)
        self.conv10 = nn.Conv2d(32, n_classes, kernel_size=1)

    def forward(self, x):
        block1 = self.block1(x)
        block2 = self.block2(block1)
        block3 = self.block3(block2)
        block4 = self.block4(block3)
        block5 = self.block5(block4)

        x = self.up_conv6(block5)
        x = torch.cat([x, block4], dim=1)
        x = self.conv6(x)

        x = self.up_conv7(x)
        x = torch.cat([x, block3], dim=1)
        x = self.conv7(x)

        x = self.up_conv8(x)
        x = torch.cat([x, block2], dim=1)
        x = self.conv8(x)

        x = self.up_conv9(x)
        x = torch.cat([x, block1], dim=1)
        x = self.conv9(x)

        x = self.up_conv10(x)
        x = self.conv10(x)

        return x

class DeepResUnet(nn.Module):
    """Deep Unet with ResNet50, ResNet101 or ResNet152 encoder."""

    def __init__(self, in_channels, n_classes=12, encoder = models.resnet101):
        super().__init__()
        self.encoder = encoder(pretrained=False)
        self.encoder.conv1 = nn.Conv2d(in_channels, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
        self.encoder_layers = list(self.encoder.children())

        self.block1 = nn.Sequential(*self.encoder_layers[:3])
        self.block2 = nn.Sequential(*self.encoder_layers[3:5])
        self.block3 = self.encoder_layers[5]
        self.block4 = self.encoder_layers[6]
        self.block5 = self.encoder_layers[7]

        self.up_conv6 = up_conv(2048, 512)
        self.conv6 = double_conv(512 + 1024, 512)
        self.up_conv7 = up_conv(512, 256)
        self.conv7 = double_conv(256 + 512, 256)
        self.up_conv8 = up_conv(256, 128)
        self.conv8 = double_conv(128 + 256, 128)
        self.up_conv9 = up_conv(128, 64)
        self.conv9 = double_conv(64 + 64, 64)
        self.up_conv10 = up_conv(64, 32)
        self.conv10 = nn.Conv2d(32, n_classes, kernel_size=1)

    def forward(self, x):
        block1 = self.block1(x)
        block2 = self.block2(block1)
        block3 = self.block3(block2)
        block4 = self.block4(block3)
        block5 = self.block5(block4)

        x = self.up_conv6(block5)
        x = torch.cat([x, block4], dim=1)
        x = self.conv6(x)

        x = self.up_conv7(x)
        x = torch.cat([x, block3], dim=1)
        x = self.conv7(x)

        x = self.up_conv8(x)
        x = torch.cat([x, block2], dim=1)
        x = self.conv8(x)

        x = self.up_conv9(x)
        x = torch.cat([x, block1], dim=1)
        x = self.conv9(x)

        x = self.up_conv10(x)
        x = self.conv10(x)

        return x

class VGGUnet(nn.Module):
    """Unet with VGG-16 or VGG-19 encoder.
    """
    def __init__(self, in_channels=1, n_classes=12, encoder=models.vgg19):
        super().__init__()
        self.encoder = encoder(pretrained=False).features
        self.encoder[0] = nn.Conv2d(in_channels, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        if encoder == models.vgg19:
            self.block1 = nn.Sequential(*self.encoder[:4])      # 64
            self.block2 = nn.Sequential(*self.encoder[4:9])     # 128
            self.block3 = nn.Sequential(*self.encoder[9:18])    # 256
            self.block4 = nn.Sequential(*self.encoder[18:27])   # 512
            self.block5 = nn.Sequential(*self.encoder[27:36])   # 512
        elif encoder == models.vgg16:
            self.block1 = nn.Sequential(*self.encoder[:4])      # 64
            self.block2 = nn.Sequential(*self.encoder[4:9])     # 128
            self.block3 = nn.Sequential(*self.encoder[9:16])    # 256
            self.block4 = nn.Sequential(*self.encoder[16:23])   # 512
            self.block5 = nn.Sequential(*self.encoder[23:30])   # 512

        self.up_conv6 = up_conv(512, 512)
        self.conv6 = double_conv(512+512, 512)
        self.up_conv7 = up_conv(512, 256)
        self.conv7 = double_conv(256+256, 256)
        self.up_conv8 = up_conv(256, 128)
        self.conv8 = double_conv(128+128, 128)
        self.up_conv9 = up_conv(128, 64)
        self.conv9 = double_conv(64+64, 64)
        self.conv10 = nn.Conv2d(64, n_classes, kernel_size=1)

    def forward(self, x):
        block1 = self.block1(x)
        block2 = self.block2(block1)
        block3 = self.block3(block2)
        block4 = self.block4(block3)
        block5 = self.block5(block4)

        x = self.up_conv6(block5)
        x = torch.cat([x, block4], dim=1)
        x = self.conv6(x)

        x = self.up_conv7(x)
        x = torch.cat([x, block3], dim=1)
        x = self.conv7(x)

        x = self.up_conv8(x)
        x = torch.cat([x, block2], dim=1)
        x = self.conv8(x)

        x = self.up_conv9(x)
        x = torch.cat([x, block1], dim=1)
        x = self.conv9(x)

        x = self.conv10(x)
        return x