DCA特征融合的pytorch代码实现

以下是一个简单的DCA特征融合的pytorch代码实现：

import torch
import torch.nn as nn

class DCA(nn.Module):
    def __init__(self, in_channels):
        super(DCA, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, in_channels, kernel_size=3, padding=1)
        self.bn1 = nn.BatchNorm2d(in_channels)
        self.relu1 = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(in_channels, in_channels, kernel_size=3, padding=1)
        self.bn2 = nn.BatchNorm2d(in_channels)
        self.relu2 = nn.ReLU(inplace=True)
        self.conv3 = nn.Conv2d(in_channels, in_channels, kernel_size=3, padding=1)
        self.bn3 = nn.BatchNorm2d(in_channels)
        self.relu3 = nn.ReLU(inplace=True)

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu1(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu2(out)

        out = self.conv3(out)
        out = self.bn3(out)

        out = out + identity
        out = self.relu3(out)

        return out

class DCA_Fusion(nn.Module):
    def __init__(self, in_channels, num_branches):
        super(DCA_Fusion, self).__init__()
        self.num_branches = num_branches
        self.dca_layers = nn.ModuleList()
        for i in range(num_branches):
            self.dca_layers.append(DCA(in_channels))

    def forward(self, x):
        out = x[0]
        for i in range(1, self.num_branches):
            out = out + x[i]
        for i in range(self.num_branches):
            x[i] = self.dca_layers[i](x[i])
        out = out / self.num_branches
        out = torch.cat(x, dim=1)
        return out

在这个实现中，我们首先定义了一个DCA模块，它包含了三个卷积层和BN层，每个卷积层后面都有ReLU激活函数。然后我们定义了一个DCA_Fusion模块，它包含了多个DCA模块，并且实现了特征融合的功能。在forward函数中，我们首先对输入的多个特征图进行求和，然后将每个特征图都输入到对应的DCA模块中进行特征增强，最后将增强后的特征图进行拼接。