AHAP: Reconstructing Arbitrary Humans from Arbitrary Perspectives with Geometric Priors
Xiaozhen Qiao, Wenjia Wang, Zhiyuan Zhao, Jiacheng Sun, Ping Luo, Hongyuan Zhang, Xuelong Li · 2026
Read AHAP: Reconstructing Arbitrary Humans from Arbitrary Perspectives with Geometric Priors by Xiaozhen Qiao, Wenjia Wang, Zhiyuan Zhao, Jiacheng Sun, Ping Luo, Hongyuan Zhang, Xuelong Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-23951.txt.
Description
Xiaozhen Qiao and six co-authors present AHAP, a feed-forward framework for rebuilding three-dimensional humans from images taken at arbitrary camera angles, without calibration. The 2026 paper notes that multi-view reconstruction usually demands checkerboards or multi-view stereo calibration, which limits scale and use in varied real scenes.
AHAP fuses multi-view geometry into association, reconstruction, and localization. A cross-view identity module matches people across views with learnable queries and soft assignment, trained by contrastive learning. A human head fuses cross-view features and scene context to predict body models, guided by reprojection losses that enforce pose consistency. Multi-view triangulation then resolves depth ambiguity that plagues single-view methods. Tests on two egocentric datasets show competitive reconstruction and camera pose accuracy at far greater speed than optimization-based rivals.
Catalog record
| CREATOR | Xiaozhen Qiao, Wenjia Wang, Zhiyuan Zhao, Jiacheng Sun, Ping Luo, Hongyuan Zhang, Xuelong Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-23951