Reproduce “Plan4D: Generative Plannable 4D Worlds” 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

Hangwei Zhang, Rujing Dang, Ruimin Lin, Jiahong Wu, Xiangxiang Chu, Chuanxia Zheng

PLAN4D predicts action-conditioned video and dense geometry from an observed scene and text action. It connects a pretrained video generator to a 3D reconstructor, reading both outputs from the same generated latent future. On 200 HoiGen examples, it cuts aligned depth error from 0.2815 to 0.1125 versus separate video generation and reconstruction, but geometry is evaluated against DA3-derived pseudo-ground truth.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.