Abstract: Current methods for large-scale point cloud scene semantic segmentation rely on manually annotated dense point-wise labels, which are costly, labor-intensive, and prone to errors.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results