1 paper · 1 filter
Bo Sun, Qixing Huang, Xiangru Huang
Existing 3D semantic segmentation methods rely on point-wise or voxel-wise feature descriptors to output segmentation predictions. However, these descriptors are often supervised a…