Reproduce “VGGT-$Ω$” 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

Jianyuan Wang, Minghao Chen, Shangzhan Zhang, Nikita Karaev, Johannes Schönberger, Patrick Labatut, Piotr Bojanowski, David Novotny and 2 more

VGGT-Ω predicts camera poses and depth maps from multiple frames in one feed-forward pass, using scene tokens to share information across views. On Sintel, its 1B model scores camera-pose AUC@3° of 35.5, versus 22.5 for MegaSaM and 16.2 for DA3-Giant. The authors noted possible overlap affecting the original checkpoint; impact is unestablished, and an overlap-checked rerun was broadly comparable.

Summary by alphaXiv.

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