1 paper
Yuanhao Su, Shaofeng Zhang
3D instance segmentation requires representations that are both geometrically structured and semantically discriminative. Pure 3D encoders model spatial structure well but often la…