SA-Net

By Qing-Long Zhang and Yu-Bin Yang

[State Key Laboratory for Novel Software Technology at Nanjing University]

This repo is the official implementation of "SA-Net: Shuffle Attention for Deep Convolutional Neural Networks".

Approach

Figure 1: The Diagram of a shuffle attention module.

Image Classification

We provide baseline sa_resnet models pretrained on ImageNet-1k.

name[email protected]#params (M)url
sa_resnet5077.8825.56BaiduDrive(474p)
sa_resnet10178.9544.55BaiduDrive(6nxm)

Evaluation

To evaluate a pre-trained sa_resnet50 on ImageNet val with a single GPU run:

python main.py -a sa_resnet50 -e --resume /path/to/sa_resnet50.pth.tar /path/to/imagenet

This should give

 * [email protected] 77.882 [email protected] 93.892

Citing SA-Net

@article{zhql2021sanet,
  title={SA-Net: Shuffle Attention for Deep Convolutional Neural Networks},
  author={Zhang, Qinglong and Yang, Yubin},
  journal={arXiv preprint arXiv:2102.00240},
  year={2021}
}
GitHub - wofmanaf/SA-Net: Code for our ICASSP 2021 paper: SA-Net: Shuffle Attention for Deep Convolutional Neural Networks
Code for our ICASSP 2021 paper: SA-Net: Shuffle Attention for Deep Convolutional Neural Networks - GitHub - wofmanaf/SA-Net: Code for our ICASSP 2021 paper: SA-Net: Shuffle Attention for Deep Convo...