Search papers, labs, and topics across Lattice.
R reconstructs large-scale
2
1
5
Achieve state-of-the-art 3D reconstruction in large-scale scenes by rapidly adapting lightweight neural networks during test time to capture global context.
LiftFormer achieves state-of-the-art monocular depth estimation by cleverly mapping image features into depth-oriented geometric subspaces, sidestepping the ill-posed nature of direct depth prediction.
