BrepGen: A B-rep Generative Diffusion Model with Structured Latent Geometry

TL;DR

BrepGen is a diffusion-based generative approach that directly outputs a Boundary representation (B-rep) Computer-Aided Design (CAD) model that incorporates free-form and doubly-curved surfaces for the first time.

Abstract

This paper presents BrepGen, a diffusion-based generative approach that directly outputs a Boundary representation (B-rep) Computer-Aided Design (CAD) model. BrepGen represents a B-rep model as a novel structured latent geometry in a hierarchical tree. With the root node representing a whole CAD solid, each element of a B-rep model (i.e., a face, an edge, or a vertex) progressively turns into a child-node from top to bottom. B-rep geometry information goes into the nodes as the global bounding box of each primitive along with a latent code describing the local geometric shape. The B-rep topology information is implicitly represented by node duplication. When two faces share an edge, the edge curve will appear twice in the tree, and a T-junction vertex with three incident edges appears six times in the tree with identical node features. 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. Results on our newly collected furniture dataset further showcase its exceptional capability in generating complicated geometry. While previous methods were limited to generating simple prismatic shapes, BrepGen incorporates free-form and doubly-curved surfaces for the first time. Additional applications of BrepGen include CAD autocomplete and design interpolation. The code, pretrained models, and dataset are available at https://github.com/samxuxiang/BrepGen.

Figures (16)

• Figure 1: Our structured latent geometry tree representation of a B-rep CAD model. Shape feature is sampled from the parametric surface or curve using grid in the parameter domain (left). Position and local geometry are extracted for the face, edge, and vertex, and then encoded as node features in the tree (mid). Topology is encoded by mating and association duplication. Node feature is represented by color values, where duplicated nodes are of the same color (right).
• Figure 2: Top row: Generation of the face bounding box position and face latent geometry. Nodes are split in half to represent each. $t$ indicates the time step and $c$ indicates the class label in the Furniture dataset. Bottom row: Generation of the edge bounding box position and edge-vertex joint latent geometry, both conditioned on the parent face. Gray color represents noisy feature values. Points are decoded from the latent geometry for visualization.
• Figure 3: Furniture B-rep Dataset overview colored by category.
• Figure 4: Statistics for the Furniture B-rep Dataset.
• Figure 5: Unconditional generation results on DeepCAD mechanical parts by (a) DeepCAD, (b) SolidGen and (c) our method BrepGen. Our method generates more realistic-looking CAD models with fewer broken geometry. Topological connections are also correct even on complicated objects.