English

MeshReGen: A Unified 3D Geometry Regeneration Framework

Computer Vision and Pattern Recognition 2026-05-19 v2

Abstract

We consider the problem of regenerating 3D objects from 2D images and initial 3D shapes. Most 3D generators operate in a one-shot fashion, converting text or images to a 3D object with limited controllability. We introduce instead MeshReGen, a 3D regenerator that is conditioned on an initial 3D shape. This conceptually simple formulation allows us to support numerous useful tasks, including 3D enhancement, reconstruction, and editing. MeshReGen uses a new conditioning mechanism based on VecSet, which allows the regenerator to update or improve the input geometry with consistent fine-grained details. MeshReGen learns a widely applicable regeneration prior from off-the-shelf 3D datasets via self-supervised pretext tasks and augmentations, without additional annotations. We evaluate both the geometric consistency and fine-grained quality of MeshReGen, achieving state-of-the-art performance in controllable 3D generation across several tasks.

Keywords

Cite

@article{arxiv.2604.28134,
  title  = {MeshReGen: A Unified 3D Geometry Regeneration Framework},
  author = {Geon Yeong Park and Roman Shapovalov and Rakesh Ranjan and Jong Chul Ye and Andrea Vedaldi and Thu Nguyen-Phuoc},
  journal= {arXiv preprint arXiv:2604.28134},
  year   = {2026}
}

Comments

Project page: https://geonyeong-park.github.io/meshregen/ 32 pages, 18 figures, 6 tables. Includes Appendix

R2 v1 2026-07-01T12:44:03.242Z