Brep2Shape: Boundary and Shape Representation Alignment via Self-Supervised Transformers

Yuanxu Sun, Yuezhou Ma, Haixu Wu, Guanyang Zeng, Muye Chen, Jianmin Wang, Mingsheng Long · 2026

Read Brep2Shape: Boundary and Shape Representation Alignment via Self-Supervised Transformers by Yuanxu Sun, Yuezhou Ma, Haixu Wu, Guanyang Zeng, Muye Chen, Jianmin Wang, Mingsheng Long (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-2602-07429.txt.

Description

Computer-aided design stores shapes as boundary representations, precise but abstract, while deep learning prefers intuitive shape data, clear but imprecise. Yuanxu Sun and six co-authors bridge that gap with Brep2Shape, a self-supervised pre-training method aligning boundary and shape representations.

A shape-aware task teaches the model to predict dense spatial points from parametric Bézier control points. A dual Transformer backbone encodes surface and curve tokens in parallel streams to respect their other shape. CAD teams get a pre-training recipe that lets networks understand physical manifolds from abstract coefficients, a step toward learning on industry-standard design files.

Catalog record

CREATORYuanxu Sun, Yuezhou Ma, Haixu Wu, Guanyang Zeng, Muye Chen, Jianmin Wang, Mingsheng Long
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE67 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-07429

Browse the full library · About the Tower of Babel