Squeeze-and-Excitation Networks（SENet）

结构和代码如下（参考：b站视频：YOLOv5 v6.1添加SE,CA,CBAM,ECA注意力机制教学，即插即用）：

Global pooling：每个channel上面的所有点做平均，这样每个channel都输出一个数。所以左图中，HxWxC变成了1x1xC。（参考：关于global average pooling理解和介绍）
FC就是全连接层。

本栏目推荐文章
• Generalised f-Mean Aggregation for Graph Neural Networks
• Graph Condensation for Graph Neural Networks
• 论文精读：ST2Vec：道路网络中的时空轨迹相似性学习（ST2Vec: Spatio_Temporal Trajectory Similarity Learning in Road Networks）
• 论文阅读-Self-supervised and Interpretable Data Cleaning with Sequence Generative Adversarial Networks
• BIgdataAIML-IBM-A neural networks deep dive - An introduction to neural networks and their programming
• Squeeze-and-Excitation Networks：SENet，早期cv中粗糙的注意力
• Relation Networks for Object Detection
• Local Relation Networks for Image Recognition： LRNet
• SeaFormer: Squeeze-enhanced Axial Transformer for Mobile Semantic Segmentation
• Fully convolutional networks for semantic segmentation