[network] channel attention network and spatial attention network

• CBAM： Address of thesis

• Purpose ：
  Convolution is to extract features by mixing the information of channel and space dimensions . In terms of attention ,SE Only pay attention to channel attention , Did not consider spatial attention . therefore , This paper proposes CBAM—— A convolution module that focuses on both channels and spatial attention , It can be used for CNNs Architecture , To improve feature map The ability to express the characteristics of .

• Network structure ：
  Main network structure
  CAM and SAM Structure
  
  CAM： The channel attention mechanism is to learn the weighting coefficient of a different channel , At the same time, all areas are considered
  
  
  SAM: Spatial attention mechanism is to learn the coefficients of different areas of the whole picture , All channels are considered at the same time .
  

• Pytorch Code implementation ：

  import torch
  from torch import nn
   
   
  class ChannelAttention(nn.Module):
      def __init__(self, in_planes, ratio=16):
          super(ChannelAttention, self).__init__()
          self.avg_pool = nn.AdaptiveAvgPool2d(1)
          self.max_pool = nn.AdaptiveMaxPool2d(1)
   
          self.fc1   = nn.Conv2d(in_planes, in_planes // 16, 1, bias=False)
          self.relu1 = nn.ReLU()
          self.fc2   = nn.Conv2d(in_planes // 16, 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=7):
          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()
          self.register_buffer()
   
      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)
  

• Reference resources ：https://blog.csdn.net/oYeZhou/article/details/116664508