1 paper
Jinsoo Park, Donggyu Choi, Ahyun Seo +2
Geometry transformers such as VGGT achieve strong performance by jointly reasoning over multiple views with global attention. However, scaling them to large view collections remain…