BRepFacetGen: Reverse Engineering B-Reps
By Generative Face Segmentation

1. Goal: Generative model for watertight B-reps
The goal of this project is to build a generative model for B-Reps. B-Reps (boundary representations) encode parametric surfaces together with precise topological structure.At inference time -- either unconditionally or conditioned on a text-prompt or a point cloud or an image -- the goal is to sample shapes represented as B-Reps that are compatible with various design, animation, and engineering software.
2. Leveraging pre-trained 3D generative models
We witnessed huge process in the last two years in generative models for 3D shapes. These models, conditioned on images or text-prompts, can sample high quality 3D shapes represented as implicit fields or meshes or some custom 3D data structure.

BibTeX

@inproceedings{hui2026brepfacetgen,
  title     = {{BRepFacetGen}: Reverse Engineering {B-Reps} By Generative Face Segmentation},
  author    = {Hui, Ka-Hei and Thamizharasan, Vikas and Jayaraman, Pradeep Kumar and Xu, Xiang},
  editor    = {Favaro, Paolo and Kukelova, Zuzana and Maki, Atsuto and Rohrbach, Anna and Schindler, Konrad and Tombari, Federico},
  booktitle = {Computer Vision -- ECCV 2026},
  year      = {2026},
  publisher = {Springer Nature Switzerland},
  address   = {Cham},
  series    = {Lecture Notes in Computer Science},
  volume    = {17054},
  pages     = {630--649},
  doi       = {10.1007/978-3-032-37258-1_35},
  url       = {https://doi.org/10.1007/978-3-032-37258-1_35},
  isbn      = {978-3-032-37258-1}
}