File size: 4,214 Bytes
231edce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import torch.nn as nn
import torch.nn.functional as F

from ...base import modules as md


class DecoderBlock(nn.Module):
    def __init__(
        self,
        in_channels,
        skip_channels,
        out_channels,
        use_batchnorm=True,
        attention_type=[None, None],
    ):
        super().__init__()
        self.conv1 = md.Conv2dReLU(
            in_channels + skip_channels,
            out_channels,
            kernel_size=3,
            padding=1,
            use_batchnorm=use_batchnorm,
        )
        self.attention1 = md.Attention(attention_type[0], in_channels=in_channels + skip_channels)
        self.conv2 = md.Conv2dReLU(
            out_channels,
            out_channels,
            kernel_size=3,
            padding=1,
            use_batchnorm=use_batchnorm,
        )
        self.attention2 = md.Attention(attention_type[1], in_channels=out_channels)

    def forward(self, x, skip=None):
        x = F.interpolate(x, scale_factor=2, mode="nearest")
        if skip is not None:
            x = torch.cat([x, skip], dim=1)
            x = self.attention1(x)
        x = self.conv1(x)
        x = self.conv2(x)
        x = self.attention2(x)
        return x


class CenterBlock(nn.Sequential):
    def __init__(self, in_channels, out_channels, use_batchnorm=True):
        conv1 = md.Conv2dReLU(
            in_channels,
            out_channels,
            kernel_size=3,
            padding=1,
            use_batchnorm=use_batchnorm,
        )
        conv2 = md.Conv2dReLU(
            out_channels,
            out_channels,
            kernel_size=3,
            padding=1,
            use_batchnorm=use_batchnorm,
        )
        super().__init__(conv1, conv2)


class UnetDecoder(nn.Module):
    def __init__(
        self,
        encoder_channels,
        decoder_channels,
        n_blocks=5,
        use_batchnorm=True,
        attention_type=None,
        center=False,
        deep_supervision=False,
    ):
        super().__init__()

        if n_blocks != len(decoder_channels):
            raise ValueError(
                "Model depth is {}, but you provide `decoder_channels` for {} blocks.".format(
                    n_blocks, len(decoder_channels)
                )
            )

        # reverse channels to start from head of encoder
        encoder_channels = encoder_channels[::-1]

        # computing blocks input and output channels
        head_channels = encoder_channels[0]
        in_channels = [head_channels] + list(decoder_channels[:-1])
        skip_channels = list(encoder_channels[1:]) + [0]
        out_channels = decoder_channels

        if center:
            self.center = CenterBlock(head_channels, head_channels, use_batchnorm=use_batchnorm)
        else:
            self.center = nn.Identity()

        self.deep_supervision = deep_supervision

        # combine decoder keyword arguments
        kwargs = dict(use_batchnorm=use_batchnorm, attention_type=[attention_type, attention_type])
        blocks = []
        for block_idx, (in_ch, skip_ch, out_ch) in enumerate(zip(in_channels, skip_channels, out_channels)):
            # For the last block, attention1 is not used
            if block_idx == (len(in_channels) - 1):
                kwargs["attention_type"] = [None, attention_type]
            blocks.append(DecoderBlock(in_ch, skip_ch, out_ch, **kwargs))
        blocks = [
            DecoderBlock(in_ch, skip_ch, out_ch, **kwargs)
            for in_ch, skip_ch, out_ch in zip(in_channels, skip_channels, out_channels)
        ]
        self.blocks = nn.ModuleList(blocks)

    def forward(self, *features):

        features = features[::-1]  # reverse channels to start from head of encoder

        head = features[0]
        skips = features[1:]

        x = self.center(head)
        
        if self.deep_supervision and self.training: outputs = []
        
        for i, decoder_block in enumerate(self.blocks):
            skip = skips[i] if i < len(skips) else None
            x = decoder_block(x, skip)
            if self.deep_supervision and self.training: outputs.append(x)

        if self.deep_supervision and self.training:
            return outputs

        return x