Search papers, labs, and topics across Lattice.
Keio University
1
0
2
Decoupling rotation and translation in positional encoding can dramatically enhance the scalability and performance of multi-view Transformers in 3D vision tasks.