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

import torch.nn as nn

class TResNetBlock(nn.Module):
    def __init__(self, in_ch, out_ch, stride=1):
        super().__init__()
        self.conv1 = nn.Conv2d(in_ch, out_ch, 3, stride, 1, bias=False)
        self.bn1 = nn.BatchNorm2d(out_ch)
        self.conv2 = nn.Conv2d(out_ch, out_ch, 3, 1, 1, bias=False)
        self.bn2 = nn.BatchNorm2d(out_ch)

        self.se = nn.Sequential(
            nn.AdaptiveAvgPool2d(1),
            nn.Conv2d(out_ch, out_ch // 16, 1),
            nn.ReLU(inplace=True),
            nn.Conv2d(out_ch // 16, out_ch, 1),
            nn.Sigmoid()
        )

        self.shortcut = nn.Sequential()
        if stride != 1 or in_ch != out_ch:
            self.shortcut = nn.Conv2d(in_ch, out_ch, 1, stride, bias=False)

    def forward(self, x):
        out = self.conv1(x).relu()
        out = self.conv2(out)
        out = out * self.se(out)
        return (out + self.shortcut(x)).relu()