RFL-CDNet: Towards accurate change detection via richer feature learning

Yuhang Gan, Wenjie Xuan, Hang Chen, Juhua Liu, Bo Du

Abstract: Highlights•We propose a novel change detection framework RFL-CDNet that utilize richer feature learning to boost change detection performance.•RFL-CDNet introduces a Coarse-To-Fine Guiding (C2FG) module to integrate the side outputs from previous coarse-scale into the current fine-scale prediction in a coarse-to-fine manner.•RFL-CDNet designs a Learnable Fusion (LF) module to fuse multiple predictions. LF module assumes that the contribution of each stage and each spatial location is independent.•RFL-CDNet sets new state-of-the-art (SOTA) performance with an F1-score of 72.28 % on WHU cultivated land dataset, 96.12 % on CDD dataset, and achieves the second best performance with F1-score of 91.39 % on WHU building dataset.