cuge1995 opened this issue 2 years ago · 1 comments
Hi, we follow PoinTr to implement 'Transformer+OcCo', replacing the encoder of PoinTr with Standard Transformer, and use the training set of ShapeNet to pretrain the Transformer.