SAUNet: Spatial-Attention Unfolding Network for Image Compressive Sensing (ACM MM 2023)

Ping Wang and Xin Yuan

• SAUNet is the first deep unfolding model for full image (NOT block) compressive sensing.
• SAUNet has achieved SOTA performance.
• More importantly, SAUNet contributes to real-world image compressive sensing systems, such as single-pixel cameras.

Proposed Method

Result

Citation

If you use SAUNet, please consider citing:

@inproceedings{wang2023saunet,
  title={SAUNet: Spatial-Attention Unfolding Network for Image Compressive Sensing},
  author={Wang, Ping and Yuan, Xin},
  booktitle={Proceedings of the 31st ACM International Conference on Multimedia},
  pages={5099--5108},
  year={2023}
}

Contact

If you have any question, please contact wangping@westlake.edu.cn

Acknowledgment: This code is based on NAFNet.