Reproduce “ShoWorld: Shape-Faithful Multi-Hypothesis Occupancy World Model” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhiqiang Yan, Guangfa Wang, Yuan Wu, Jian Yang, Gim Hee Lee
ShoWorld forecasts future 3D occupancy by representing each moving object’s motion as probability-weighted trajectory hypotheses and its shape as sub-voxel points. Under ground-truth occupancy and ego trajectory on Occ3D-nuScenes, it averages 53.96% mIoU and 65.40% IoU across 1-, 2-, and 3-second horizons, versus GAST’s 47.40% and 56.24%. These benchmark results do not establish equivalent performance with sensor-estimated inputs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.