【YOLO改进】主干插入S2Attention模块(基于MMYOLO)

最新推荐文章于 2025-07-21 15:19:03 发布

五山一胖

最新推荐文章于 2025-07-21 15:19:03 发布

阅读量788

点赞数 18

CC 4.0 BY-SA版权

分类专栏： YOLO改进文章标签： YOLO 深度学习计算机视觉

本文链接：https://blog.youkuaiyun.com/Vlone_pp/article/details/137524138

S2Attention模块

论文链接: https://arxiv.org/abs/2108.01072

将S2Attention模块添加到MMYOLO中

将开源代码S2Attention.py文件复制到mmyolo/models/plugins目录下
导入MMYOLO用于注册模块的包: from mmyolo.registry import MODELS
确保 class S2Attention中的输入维度为in_channels(因为MMYOLO会提前传入输入维度参数，所以要保持参数名的一致)
利用@MODELS.register_module()将“class S2Attention(nn.Module)”注册:
修改mmyolo/models/plugins/__init__.py文件
在终端运行:
```
python setup.py install
```
修改对应的配置文件，并且将plugins的参数“type”设置为“S2Attention”，可参考【YOLO改进】主干插入注意力机制模块CBAM（基于MMYOLO）-优快云博客

修改后的S2Attention.py

import torch
from torch import nn
from mmyolo.registry import MODELS

def spatial_shift1(x):
    b, w, h, c = x.size()
    x[:, 1:, :, :c // 4] = x[:, :w - 1, :, :c // 4]
    x[:, :w - 1, :, c // 4:c // 2] = x[:, 1:, :, c // 4:c // 2]
    x[:, :, 1:, c // 2:c * 3 // 4] = x[:, :, :h - 1, c // 2:c * 3 // 4]
    x[:, :, :h - 1, 3 * c // 4:] = x[:, :, 1:, 3 * c // 4:]
    return x


def spatial_shift2(x):
    b, w, h, c = x.size()
    x[:, :, 1:, :c // 4] = x[:, :, :h - 1, :c // 4]
    x[:, :, :h - 1, c // 4:c // 2] = x[:, :, 1:, c // 4:c // 2]
    x[:, 1:, :, c // 2:c * 3 // 4] = x[:, :w - 1, :, c // 2:c * 3 // 4]
    x[:, :w - 1, :, 3 * c // 4:] = x[:, 1:, :, 3 * c // 4:]
    return x


class SplitAttention(nn.Module):
    def __init__(self, channel=512, k=3):
        super().__init__()
        self.channel = channel
        self.k = k
        self.mlp1 = nn.Linear(channel, channel, bias=False)
        self.gelu = nn.GELU()
        self.mlp2 = nn.Linear(channel, channel * k, bias=False)
        self.softmax = nn.Softmax(1)

    def forward(self, x_all):
        b, k, h, w, c = x_all.shape
        x_all = x_all.reshape(b, k, -1, c)  # bs,k,n,c
        a = torch.sum(torch.sum(x_all, 1), 1)  # bs,c
        hat_a = self.mlp2(self.gelu(self.mlp1(a)))  # bs,kc
        hat_a = hat_a.reshape(b, self.k, c)  # bs,k,c
        bar_a = self.softmax(hat_a)  # bs,k,c
        attention = bar_a.unsqueeze(-2)  # #bs,k,1,c
        out = attention * x_all  # #bs,k,n,c
        out = torch.sum(out, 1).reshape(b, h, w, c)
        return out

@MODELS.register_module()
class S2Attention(nn.Module):

    def __init__(self, in_channels=512):
        super().__init__()
        self.mlp1 = nn.Linear(in_channels, in_channels * 3)
        self.mlp2 = nn.Linear(in_channels, in_channels)
        self.split_attention = SplitAttention()

    def forward(self, x):
        b, c, w, h = x.size()
        x = x.permute(0, 2, 3, 1)