Reproduce “ObjectWorld: Scaling Object-Centric World Models with A Large-Scale 3D Full-Object Flow Dataset” 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
Haonan Chen, Sicheng Isabella Lu, Haitian Li, Jiajin Wu, Hexiang Wei, Jiawei Zhang, Xin Fei, Yiwen Hou and 4 more
ObjectWorld predicts 3D trajectories for points across a reconstructed object’s full surface from an RGB-D observation and task instruction, then fits robot motion to that target. Its dataset contains 1,210.1 hours of simulated and real video; against partial-flow annotations, it lowers median normalized endpoint error and increases surface coverage. Real-world trials show viability, with only 20 trials per method.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.