Return
Single stage weakly supervised semantic segmentation via enhanced patch affinity
DOI:10.1016/j.imavis.2025.105791.png)
Abstract
En 中文
• Single stage WSSS method that integrates patch affinity. • Adaptive evaluation of attention importance from different layers and weighted fusion. • Head averaging can reduce occupancy of attention map and maintain optimization. • Experimental results have shown that our method achieves competitive performance.
Journal
IF:
4.2
Papers:
4.1K
Citations:
6.7K
Organization
Cited Papers
No cited papers available

