BrepGen represents a CAD B-rep as a novel structured latent geometry hierarchical tree, with each primitive (face, edge, or vertex) progressively turns into a child-node from top to bottom. The B-rep geometry is encoded into the tree nodes along with a latent code describing the local geometric shape. The B-rep topology information is implicitly represented by node duplication.
Starting from the root and progressing to the leaf, BrepGen employs Transformer-based diffusion models to sequentially denoise node features while duplicated nodes are detected and merged, recovering the B-Rep topology information. Extensive experiments show that BrepGen advances the task of CAD B-rep generation, surpassing existing methods on various benchmarks.
@article{xu2024brepgen,
title={Brepgen: A b-rep generative diffusion model with structured latent geometry},
author={Xu, Xiang and Lambourne, Joseph and Jayaraman, Pradeep and Wang, Zhengqing and Willis, Karl and Furukawa, Yasutaka},
journal={ACM Transactions on Graphics (TOG)},
volume={43},
number={4},
pages={1--14},
year={2024},
publisher={ACM New York, NY, USA}
}