arxivcs.CV2026-07-21
Learning Semantic-Robust Change Detection via Semantic-Invariant Self-Distillation
Jiuhe Qu, Yingping Liang, Ying Fu
Change detection aims to identify semantic changes between remote sensing images. However, features from models are easily disturbed by non-semantic variations, such as illumination, shadows, and atmospheric changes, leading to false alarms and limited generalization in real-worl…