Reproduce “GPA-VGGT: Geometry-Preserving Acceleration of VGGT” 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

Muyeol Sung, Jangwon Kim, Jongwoo Lim

GPA-VGGT accelerates multi-view 3D perception by reducing the keys and values used in attention while preserving each token’s query and retaining compact shared cross-frame context in early blocks. On CO3D, it scores 56.94 AUC@3 versus 57.49 for vanilla VGGT, and runs 2.58 times faster on 200 frames; these results are limited to tested settings.

Summary by alphaXiv.

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