U3M: Unbiased Multiscale Modal Fusion Model for Multimodal Semantic Segmentation

Bingyu Li, Da Zhang, Zhiyuan Zhao, Junyu Gao, Xuelong Li

Abstract: Highlights•Introduce U3M, a novel unbiased multiscale modal fusion model.•Design a novel multiscale model fusion layer.•Sufficient experiments are conducted to validate the effectiveness of U3M.•Demonstrated state-of-the-art results on multiple challenging datasets.•Enabled autonomous generation of modality preferences for varied contexts.