Abstract: Efficient semantic segmentation of large-scale point cloud scenes is regarded as a fundamental and essential task for perceiving and understanding 3-D environments. It is also recognized as ...