YOLOv5 加入SE注意力机制

方法1：

步骤1：在common.py中添加SE模块

class SE(nn.Module):
    def __init__(self, c1, c2, r=16):
        super(SE, self).__init__()
        self.avgpool = nn.AdaptiveAvgPool2d(1)
        self.l1 = nn.Linear(c1, c1 // r, bias=False)
        self.relu = nn.ReLU(inplace=True)
        self.l2 = nn.Linear(c1 // r, c1, bias=False)
        self.sig = nn.Sigmoid()
    def forward(self, x):
        print(x.size())
        b, c, _, _ = x.size()
        y = self.avgpool(x).view(b, c)
        y = self.l1(y)
        y = self.relu(y)
        y = self.l2(y)
        y = self.sig(y)
        y = y.view(b, c, 1, 1)
        return x * y.expand_as(x)

 步骤2：在yolo.py文件中加入类名

 新建yolos_se.yaml

注意力机制可以加在Backbone、Neck、Head等部分，常见的有两种：一种是在主干的SPPF前面添加一层；二是将Backbone中的C3全部替换。 

yolo.py

 

 运行yolo.py

 修改train.py中的'--cfg'默认参数

 

方法2： 

model下创建SE.py

import numpy as np
import torch
from torch import nn
from torch.nn import init
 
class SEAttention(nn.Module):
 
    def __init__(self, channel=512,reduction=16):
        super().__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.fc = nn.Sequential(
            nn.Linear(channel, channel // reduction, bias=False),
            nn.ReLU(inplace=True),
            nn.Linear(channel // reduction, channel, bias=False),
            nn.Sigmoid()
        )
 
 
    def init_weights(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                init.kaiming_normal_(m.weight, mode='fan_out')
                if m.bias is not None:
                    init.constant_(m.bias, 0)
            elif isinstance(m, nn.BatchNorm2d):
                init.constant_(m.weight, 1)
                init.constant_(m.bias, 0)
            elif isinstance(m, nn.Linear):
                init.normal_(m.weight, std=0.001)
                if m.bias is not None:
                    init.constant_(m.bias, 0)
 
    def forward(self, x):
        b, c, _, _ = x.size()
        y = self.avg_pool(x).view(b, c)
        y = self.fc(y).view(b, c, 1, 1)
        return x * y.expand_as(x)

 yolo.py下导入SE模块

from models.SE import SEAttention

并加入下面代码

修改配置文件

 结果：

添加成功