Search papers, labs, and topics across Lattice.
ETH Zurich, Switzerland Delft University of Technology, Netherlands Microsoft, Switzerland
Microsoft Research2
0
4
13
ReViV reconstructs 4D viewer and view dynamics from a single monocular video, achieving unprecedented accuracy and speed without heavy task-specific priors.
Forget expensive 3D training data: Loc3R-VLM shows how to give 2D vision-language models strong 3D spatial reasoning by distilling knowledge from a pretrained 3D foundation model using only monocular video.
