Split-and-Fit: Learning B-Reps via Structure-Aware Voronoi Partitioning

Yilin Liu,Jiale Chen,Shanshan Pan,Daniel Cohen-Or,Hao Zhang,Hui Huang

2024-06-08

Abstract:We introduce a novel method for acquiring boundary representations (B-Reps) of 3D CAD models which involves a two-step process: it first applies a spatial partitioning, referred to as the ``split``, followed by a ``fit`` operation to derive a single primitive within each partition. Specifically, our partitioning aims to produce the classical Voronoi diagram of the set of ground-truth (GT) B-Rep primitives. In contrast to prior B-Rep constructions which were bottom-up, either via direct primitive fitting or point clustering, our Split-and-Fit approach is top-down and structure-aware, since a Voronoi partition explicitly reveals both the number of and the connections between the primitives. We design a neural network to predict the Voronoi diagram from an input point cloud or distance field via a binary classification. We show that our network, coined NVD-Net for neural Voronoi diagrams, can effectively learn Voronoi partitions for CAD models from training data and exhibits superior generalization capabilities. Extensive experiments and evaluation demonstrate that the resulting B-Reps, consisting of parametric surfaces, curves, and vertices, are more plausible than those obtained by existing alternatives, with significant improvements in reconstruction quality. Code will be released on <a class="link-external link-https" href="https://github.com/yilinliu77/NVDNet" rel="external noopener nofollow">this https URL</a>.

Computer Vision and Pattern Recognition,Graphics

What problem does this paper attempt to address?

This paper proposes a new method called Split-and-Fit for learning the boundary representation (B-Rep) of 3D CAD models through structurally-aware Voronoi partitioning. Traditional B-Rep construction methods are usually bottom-up, relying on point clustering or primitive fitting. Split-and-Fit, on the other hand, is top-down and considers structural information as it reveals the number and connections of basic shapes using Voronoi diagrams.

Specifically, this method first partitions the volumetric space ("split") and then fits a single primitive shape in each partition. This is achieved by training a neural network called NVD-Net to predict the Voronoi diagram of input point clouds or distance fields. This approach reduces the reliance on local geometric or topological features and directly optimizes global structural properties such as the number of primitives.

Experiments show that compared to existing methods, Split-and-Fit produces more reasonable B-Rep reconstructions with lower geometric errors, higher topological consistency, better instance recognition performance, and improved generalization capability. The code will be released on GitHub.