1 paper
Shaochen Zhang, Zekun Qi, Runpei Dong +2
We rethink the role of positional encoding in 3D representation learning and fine-tuning. We argue that using positional encoding in point Transformer-based methods serves to aggre…