Reproduce “Top2Pano: Learning to Generate Indoor Panoramas from Top-Down View” 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

Zitong Zhang, Suranjan Gautam, Rui Yu

Top2Pano generates indoor panoramas from top-down views by estimating a 3D occupancy map. Given a camera position, it renders coarse color and depth panoramas to guide a diffusion model’s visual refinement. On Matterport3D, its FID is 30.84 versus PanFusion’s 85.74; it also leads listed baselines on four metrics across both datasets, but this does not establish accurate geometry.

Summary by alphaXiv.

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