Reproduce “World-delta: Continually Integrating Observations for Incremental 3D Scene Generation” 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 Wang, Hanyu Zhou, Xiaobin Hu, Luxin Yan

World-δ maintains a 3D scene by distinguishing observation-supported content from provisional completions. It compares incoming RGB evidence, using geometric and semantic cues, with the state, then keeps, combines, corrects, or extends affected regions. On ScanNet-v2, it reports PSNR 28.7, SSIM 0.841, LPIPS 0.192, and 6.1 seconds per update versus FlexWorld’s 24.3, 0.776, 0.289, and 35 minutes; it assumes static scenes.

Summary by alphaXiv.

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