LPA.py

import torch
import torch.nn as nn
#论文：SwinPA-Net: Swin Transformer-Based Multiscale Feature Pyramid Aggregation Network for Medical Image Segmentation
#论文地址：https://ieeexplore.ieee.org/document/9895210

class ChannelAttention(nn.Module):
    def __init__(self, in_planes):
        super(ChannelAttention, self).__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.max_pool = nn.AdaptiveMaxPool2d(1)

        self.fc1 = nn.Conv2d(in_planes, in_planes // 8, 1, bias=False)
        self.relu1 = nn.ReLU()
        self.fc2 = nn.Conv2d(in_planes // 8, in_planes, 1, bias=False)

        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        avg_out = self.fc2(self.relu1(self.fc1(self.avg_pool(x))))
        max_out = self.fc2(self.relu1(self.fc1(self.max_pool(x))))
        out = avg_out + max_out
        return self.sigmoid(out)


class SpatialAttention(nn.Module):
    def __init__(self, kernel_size=3):
        super(SpatialAttention, self).__init__()

        assert kernel_size in (3, 7), 'kernel size must be 3 or 7'
        padding = 3 if kernel_size == 7 else 1

        self.conv1 = nn.Conv2d(2, 1, kernel_size, padding=padding, bias=False)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        avg_out = torch.mean(x, dim=1, keepdim=True)
        max_out, _ = torch.max(x, dim=1, keepdim=True)
        x = torch.cat([avg_out, max_out], dim=1)
        x = self.conv1(x)
        return self.sigmoid(x)


class LPA(nn.Module):
    def __init__(self, in_channel):
        super(LPA, self).__init__()
        self.ca = ChannelAttention(in_channel)
        self.sa = SpatialAttention()

    def forward(self, x):
        x0, x1 = x.chunk(2, dim=2)
        x0 = x0.chunk(2, dim=3)
        x1 = x1.chunk(2, dim=3)
        x0 = [self.ca(x0[-2]) * x0[-2], self.ca(x0[-1]) * x0[-1]]
        x0 = [self.sa(x0[-2]) * x0[-2], self.sa(x0[-1]) * x0[-1]]

        x1 = [self.ca(x1[-2]) * x1[-2], self.ca(x1[-1]) * x1[-1]]
        x1 = [self.sa(x1[-2]) * x1[-2], self.sa(x1[-1]) * x1[-1]]

        x0 = torch.cat(x0, dim=3)
        x1 = torch.cat(x1, dim=3)
        x3 = torch.cat((x0, x1), dim=2)

        x4 = self.ca(x) * x
        x4 = self.sa(x4) * x4
        x = x3 + x4
        return x


if __name__ == '__main__':

    input = torch.rand(1, 28, 64, 64)
    block = LPA(in_channel=28)
    output = block(input)

    print(input.size())
    print(output.size())