yolov添加SE注意力机制

YOLOv3是一种目标检测模型，它使用卷积神经网络来检测图像中的物体。SE( Squeeze-and-Excitation)注意力机制是一种提高模型表现的技术，它可以自适应地学习通道间的重要性，并对这些通道进行加权。

要将SE注意力机制添加到YOLOv3中，需要在YOLOv3的卷积层之后添加SE块。SE块由两个部分组成：压缩部分和激励部分。压缩部分通过全局平均池化将通道压缩到一维，然后通过两个全连接层将其映射到一个较小的向量。激励部分对该向量进行Sigmoid激活，然后将其与卷积特征图相乘，以得到加权特征图。

在YOLOv3中，可以通过以下方式实现SE注意力机制的添加：

1. 定义SE块的压缩和激励部分。

import torch.nn as nn

class SEBlock(nn.Module):
    def __init__(self, in_channels, reduction_ratio=16):
        super(SEBlock, self).__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.fc1 = nn.Linear(in_channels, in_channels // reduction_ratio)
        self.fc2 = nn.Linear(in_channels // reduction_ratio, in_channels)
        self.relu = nn.ReLU(inplace=True)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        b, c, _, _ = x.size()
        y = self.avg_pool(x).view(b, c)
        y = self.fc1(y)
        y = self.relu(y)
        y = self.fc2(y)
        y = self.sigmoid(y).view(b, c, 1, 1)
        return x * y

2. 在YOLOv3的卷积层之后添加SE块。

import torch.nn as nn

class YOLOv3(nn.Module):
    def __init__(self, num_classes=80):
        super(YOLOv3, self).__init__()
        ...
        self.conv1 = nn.Conv2d(512, 1024, kernel_size=3, stride=1, padding=1)
        self.bn1 = nn.BatchNorm2d(1024)
        self.se1 = SEBlock(1024) # 添加SE块
        self.conv2 = nn.Conv2d(1024, 512, kernel_size=1, stride=1, padding=0)
        self.bn2 = nn.BatchNorm2d(512)
        self.conv3 = nn.Conv2d(512, 1024, kernel_size=3, stride=1, padding=1)
        self.bn3 = nn.BatchNorm2d(1024)
        self.se2 = SEBlock(1024) # 添加SE块
        self.conv4 = nn.Conv2d(1024, 512, kernel_size=1, stride=1, padding=0)
        self.bn4 = nn.BatchNorm2d(512)
        self.conv5 = nn.Conv2d(512, 1024, kernel_size=3, stride=1, padding=1)
        self.bn5 = nn.BatchNorm2d(1024)
        self.se3 = SEBlock(1024) # 添加SE块
        self.conv6 = nn.Conv2d(1024, 512, kernel_size=1, stride=1, padding=0)
        self.bn6 = nn.BatchNorm2d(512)
        self.conv7 = nn.Conv2d(512, 1024, kernel_size=3, stride=1, padding=1)
        self.bn7 = nn.BatchNorm2d(1024)
        self.se4 = SEBlock(1024) # 添加SE块
        self.conv8 = nn.Conv2d(1024, 512, kernel_size=1, stride=1, padding=0)
        self.bn8 = nn.BatchNorm2d(512)
        self.conv9 = nn.Conv2d(512, 1024, kernel_size=3, stride=1, padding=1)
        self.bn9 = nn.BatchNorm2d(1024)
        self.se5 = SEBlock(1024) # 添加SE块
        ...
        
    def forward(self, x):
        ...
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.se1(x) # 添加SE块
        x = self.relu(x)
        x = self.conv2(x)
        x = self.bn2(x)
        x = self.relu(x)
        x = self.conv3(x)
        x = self.bn3(x)
        x = self.se2(x) # 添加SE块
        x = self.relu(x)
        x = self.conv4(x)
        x = self.bn4(x)
        x = self.relu(x)
        x = self.conv5(x)
        x = self.bn5(x)
        x = self.se3(x) # 添加SE块
        x = self.relu(x)
        x = self.conv6(x)
        x = self.bn6(x)
        x = self.relu(x)
        x = self.conv7(x)
        x = self.bn7(x)
        x = self.se4(x) # 添加SE块
        x = self.relu(x)
        x = self.conv8(x)
        x = self.bn8(x)
        x = self.relu(x)
        x = self.conv9(x)
        x = self.bn9(x)
        x = self.se5(x) # 添加SE块
        x = self.relu(x)
        ...
        return x

这样，我们就成功地将SE注意力机制添加到了YOLOv3中。