Reproduce “DIRT: Depth-completed Image-based Rendering using Two-band Blending for Live Novel-view Synthesis” 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

Mathias Harrer, Laura Fink, Noah Lewis, Paul Himmler, Tim Weyrich

DIRT produces live novel views from posed images and measured depth, using a completed target-view depth map to warp and blend captured images without scene-specific optimization. On Argoverse 2, it reports PSNR 25.05, LPIPS 0.1802, and 135 ms, versus On-the-fly NVS at 18.19, 0.5474, and 486 ms. Tests evaluate next viewpoints from the preceding five seconds in limited-dynamic automotive scenes.

Summary by alphaXiv.

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