Yolov5 improvement 4: add ECA channel attention mechanism

  front       said ： As the current advanced deep learning target detection algorithm YOLOv5, A large number of trick, But when dealing with some complex background problems , It is still prone to the problem of mistakes and omissions . Subsequent articles , Focus on YOLOv5 How to improve is introduced in detail , The purpose is to provide their own meager help and reference for those who need innovation in scientific research or friends who need to achieve better results in engineering projects .

solve the problem ： Join in ECA Attention mechanism , It can make the network pay more attention to the target to be detected , Improve the detection effect . The attention mechanism is an article CVPR2020 Articles on improving channel attention ---ECANet,ECANet Mainly for SENet Some improvements have been made to the module , Put forward a kind of Local cross channel interaction strategy without dimensionality reduction （ECA modular ） And adaptive selection of one-dimensional convolution kernel size , So the performance is improved .

Add method ：

First step ： Determine where to add , As a plug and play attention module , Can be added to YOLOv5 Anywhere in the network .

The second step ：common.py structure  Eca_layer modular .

class Eca_layer(nn.Module):
    """Constructs a ECA module.

    Args:
        channel: Number of channels of the input feature map
        k_size: Adaptive selection of kernel size
    """
    def __init__(self, channel, k_size=3):
        super(Eca_layer, self).__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.conv = nn.Conv1d(1, 1, kernel_size=k_size, padding=(k_size - 1) // 2, bias=False)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        # x: input features with shape [b, c, h, w]
        b, c, h, w = x.size()

        # feature descriptor on the global spatial information
        y = self.avg_pool(x)

        # Two different branches of ECA module
        y = self.conv(y.squeeze(-1).transpose(-1, -2)).transpose(-1, -2).unsqueeze(-1)

        # Multi-scale information fusion
        y = self.sigmoid(y)

        return x * y.expand_as(x)

The third step ：yolo.py Register in Eca_layer modular

        elif m is Eca_layer:
            channel,k_size=args[0],args[1]
            channel = make_divisible(channel * gw, 8) if channel != no else channel
            args = [channel, k_size]

Step four ： modify yaml file , This paper aims to modify backbone For example , The original C3 Add the module after the module .

 

Step five ： take train.py In this article yaml File can , Start training .

  junction     fruit ： I have done a lot of experiments on multiple data sets , For different data sets, the effect is different , There are also differences in the methods of adding locations to the same dataset , You need to experiment . Most cases are effective and improved .

Let me know ： The next article shares how to modify the feature fusion network structure , The original BIFPN Change it to BIFPN, Improve the feature fusion ability of the algorithm . Interested friends can pay attention to me , If you have questions, you can leave a message or chat with me in private

PS：ECA Attention mechanism , Not only can it be added YOLOv5, It can also be added to any other deep learning network , Whether it is classification, detection or segmentation , Mainly in the field of computer vision , May have different degrees of improvement effect .

Last , I hope I can powder each other , Be a friend , Learn and communicate together .