← back to paper
arxiv: 2604.02546 · 2 revisions
RGB-Pointmap Pretraining for Unified 3D Scene Understanding