Semi-supervised semantic segmentation with confidence-driven consistency learning

doi:10.1016/j.eswa.2025.128965

Semi-supervised semantic segmentation with confidence-driven consistency learning

2025-07-09

0

PRE

AI

DOI：10.1016/j.eswa.2025.128965

Original

Original request for help

Save

Abstract

En 中文

• The CDCL is proposed to fully exploit the information in unlabeled data. • The RPR strategy reduces inter-class competition and enhances prediction certainty. • The NLA module is designed to extract robust information from low-confidence pixels. • Experimental results on two benchmark datasets demonstrate the superiority of CDCL.

Semi-supervised semantic segmentation with confidence-driven consistency learning

Abstract

Journal

Organization