BriefGPT.xyz
Feb, 2024
语义分割的交换分配弱监督共训练
Weakly Supervised Co-training with Swapping Assignments for Semantic Segmentation
HTML
PDF
Xinyu Yang, Hossein Rahmani, Sue Black, Bryan M. Williams
TL;DR
提出了一种采用引导类激活图的端到端弱监督语义分割模型CoSA,通过交替学习和三种技术的引入,有效减少了类激活图的不一致性和错误,达到了优于现有方法的性能。
Abstract
class activation maps
(CAMs) are commonly employed in
weakly supervised semantic segmentation
(WSSS) to produce pseudo-labels. Due to incomplete or excessive class activation, existing studies often resort to off
→