Abstract: High-precision semantic segmentation of multimodal remote sensing imagery is critical for applications such as urban planning and land monitoring. However, existing methods often suffer from ...