Search papers, labs, and topics across Lattice.
2
0
3
A massive 2.6M-scene benchmark and a novel generative model finally unlock robust dexterous grasping in cluttered environments, outperforming prior methods without post-optimization.
DreamHand achieves a groundbreaking 40% reduction in error for 3D hand trajectory recovery in occluded environments, setting a new standard for egocentric video analysis.
