目录

 1.GCT介绍

实验结果

 2.GCT引入到yolov5

2.1 加入common.py中：

2.2 加入yolo.py中：

2.3 yolov5s_GCT.yaml

2.4 yolov5s_GCT1.yaml

---

 1.GCT介绍

 论文：https://openaccess.thecvf.com/content/CVPR2021/papers/Ruan_Gaussian_Context_Transformer_CVPR_2021_paper.pdf

浙江大学等机构发布的一篇收录于CVPR2021的文章，提出了一种新的通道注意力结构，在几乎不引入参数的前提下优于大多SOTA通道注意力模型，如SE、ECA等。这篇文章虽然叫Gaussian Context Transformer，但是和Transformer并无太多联系，这里可以理解为高斯上下文变换器。

         LCT（linear context transform）观察所得，如下图所示，SE倾向于学习一种负相关，即全局上下文偏离均值越多，得到的注意力激活值就越小。为了更加精准地学习这种相关性，LCT使用一个逐通道地变换来替代SE中的两个全连接层。然而，实验表明，LCT学得的这种负相关质量并不是很高，下图中右侧可以看出，LCT的注意力激活值波动是很大的。
 

         在本文中，我们假设这种关系是预先确定的。基于这个假设，我们提出了一个简单但极其有效的通道注意力块，称为高斯上下文Transformer (GCT)，它使用满足预设关系的高斯函数实现上下文特征激励。

实验结果

在ImageNet 和 MS COCO 基准测试的大量实验表明，我们的 GCT 导致各种深度 CNN 和检测器的持续改进。与一系列最先进的通道注意力块（例如 SE 和 ECA）相比，我们的 GCT 在有效性和效率方面更为出色。

 2.GCT引入到yolov5

2.1 加入common.py中：

###################### Gaussian Context Transformer  attention  ####     END   by  AI&CV  ###############################

""" 
PyTorch implementation of Gaussian Context Transformer

As described in http://openaccess.thecvf.com//content/CVPR2021/papers/Ruan_Gaussian_Context_Transformer_CVPR_2021_paper.pdf

Gaussian Context Transformer (GCT), which achieves contextual feature excitation using
a Gaussian function that satisfies the presupposed relationship.
"""

import torch
from torch import nn


class GCT(nn.Module):
    def __init__(self, channels, c=2, eps=1e-5):
        super().__init__()
        self.avgpool = nn.AdaptiveAvgPool2d(1)
        self.eps = eps
        self.c = c

    def forward(self, x):
        y = self.avgpool(x)
        mean = y.mean(dim=1, keepdim=True)
        mean_x2 = (y ** 2).mean(dim=1, keepdim=True)
        var = mean_x2 - mean ** 2
        y_norm = (y - mean) / torch.sqrt(var + self.eps)
        y_transform = torch.exp(-(y_norm ** 2 / 2 * self.c))
        return x * y_transform.expand_as(x)


###################### Gaussian Context Transformer  attention  ####     END   by  AI&CV  ###############################

2.2 加入yolo.py中：

def parse_model(d, ch): # model_dict, input_channels(3)

添加以下内容 

 if m in {Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF,DWConv, MixConv2d, Focus, CrossConv,BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, CNeB, nn.ConvTranspose2d, DWConvTranspose2d, C3x, C2f,GCT}:

2.3 yolov5s_GCT.yaml

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 1  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # layer channel multiple
anchors:
  - [10,13, 16,30, 33,23]  # P3/8
  - [30,61, 62,45, 59,119]  # P4/16
  - [116,90, 156,198, 373,326]  # P5/32

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3, [128]],
   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3, [256]],
   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3, [512]],
   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3, [1024]],
   [-1, 1, SPPF, [1024, 5]],  # 9
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3, [512, False]],  # 13

   [-1, 1, Conv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)

   [-1, 1, Conv, [256, 3, 2]],
   [[-1, 14], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)

   [-1, 1, Conv, [512, 3, 2]],
   [[-1, 10], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
   [-1, 1, GCT, [1024]],  # 24

   [[17, 20, 24], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

2.4 yolov5s_GCT1.yaml

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 1  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # layer channel multiple
anchors:
  - [10,13, 16,30, 33,23]  # P3/8
  - [30,61, 62,45, 59,119]  # P4/16
  - [116,90, 156,198, 373,326]  # P5/32

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3, [128]],
   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3, [256]],
   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3, [512]],
   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3, [1024]],
   [-1, 1, SPPF, [1024, 5]],  # 9
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3, [512, False]],  # 13

   [-1, 1, Conv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
   [-1, 1, GCT, [256]],  # 18

   [-1, 1, Conv, [256, 3, 2]],
   [[-1, 14], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3, [512, False]],  # 21 (P4/16-medium)
   [-1, 1, GCT, [512]],  # 22

   [-1, 1, Conv, [512, 3, 2]],
   [[-1, 10], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3, [1024, False]],  # 25 (P5/32-large)
   [-1, 1, GCT, [1024]],  # 26

   [[18, 22, 26], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

3.YOLOv5/YOLOv7魔术师专栏介绍
 

💡💡💡YOLOv5/YOLOv7魔术师，独家首发创新（原创），持续更新，最终完结篇数≥100+，适用于Yolov5、Yolov7、Yolov8等各个Yolo系列，专栏文章提供每一步步骤和源码，轻松带你上手魔改网络
💡💡💡重点：通过本专栏的阅读，后续你也可以自己魔改网络，在网络不同位置（Backbone、head、detect、loss等）进行魔改，实现创新！！！
 专栏介绍：
✨✨✨原创魔改网络、复现前沿论文，组合优化创新
🚀🚀🚀小目标、遮挡物、难样本性能提升
🍉🍉🍉持续更新中，定期更新不同数据集涨点情况
本专栏提供每一步改进步骤和源码，开箱即用，在你的数据集下轻松涨点
通过注意力机制、小目标检测、Backbone&Head优化、 IOU&Loss优化、优化器改进、卷积变体改进、轻量级网络结合yolo等方面进行展开点
专栏链接如下：
https://blog.csdn.net/m0_63774211/category_12240482.html