Reproduce “Sat2Scene: 3D Urban Scene Generation from Satellite Images with Diffusion” 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
Zuoyue Li, Zhenqiang Li, Zhaopeng Cui, Marc Pollefeys, Martin R. Oswald
Sat2Scene generates a shared 3D scene from satellite-inferred geometry and renders views from poses. It uses sparse 3D diffusion to color scene points, encodes features at those points, and applies a neural renderer. On HoliCity, it improves video-distribution and overlap-region consistency scores over MVDiffusion, but trails it on single-frame FID and KID; satellite-to-ground results on OmniCity are qualitative only.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.