PRX-Change: Enhancing remote sensing change detection through progressive feature refinement and Cross-Attention interaction

Haotian Tan, Linchao He, Wenchao Du, Hong Liu, Hu Chen, Yi Zhang, Hongyu Yang

Abstract: Highlights•We proposed a generalized change detection (CD) framework PRX-Change, which gains SOTA performance on multiple remote sensing CD benchmarks.•We introduced a novel feature refinement method termed PFRM to refine hard features.•We designed X-Cross, an enhanced Cross-Attention module for equitable feature interaction. It adapts better to CD tasks compared to conventional approaches.