NTU's HiP3R Enables Consistent 3D Reconstruction for Long Videos
NTU researchers release HiP3R, a hierarchical pose graph approach that stitches local 3D reconstructions to keep long videos of thousands of frames globally consistent, cutting average ATE from 8.29m to 4.02m.
2026-10-24 ~ 2026-10-25 · 2 related posts
- Episode 1: NTU's HiP3R Enables Consistent 3D Reconstruction for Long Videos(2026-10-24, 2 posts)
- Episode 2: HiP3R Open-Sourced: Reconstructing Thousands-Frame Videos into Consistent 3D Scenes(2026-10-26, 3 posts)
- HiP3R: hierarchical pose graphs cut long-horizon 3D reconstruction ATE from 8.3m to 4.0m — zhenjun_zhao · 2026-10-24
- HiP3R stitches local 3D reconstructions for consistent scenes across thousands of video frames — ccloy · 2026-10-25