Search papers, labs, and topics across Lattice.
Yonsei University
1
0
3
1
Small, fast-moving objects and fine textures in dynamic volumetric video are finally recoverable by intelligently allocating Gaussians based on future motion, not just current positions.