Neural networks learn stable 3D solid shapes from canonical CAD inputs
Learn the Solid, Not the File: Canonical Inputs for Neural Networks on CAD Boundary Representations
Computer Vision and Pattern RecognitionArtificial IntelligenceComputational Geometry
Summary
3D models made by engineers can be saved in many different ways even when showing the same object, which confuses computer programs trying to understand them. The authors show that current AI methods fail to recognize the same shape when its description changes. They created a new way to describe 3D objects so the computer always sees the real shape, not the differences in files. Their method is more reliable for recognizing parts, even when the files vary or the object is moved around.
What this means in practice
- •For cad software developers: Build more robust AI tools that understand 3D parts consistently despite internal file variations in CAD models.
- •For manufacturing engineers: Use more stable AI interpretations of CAD models for quality control and defect detection in designed parts.
Authors
Heinrich Jiang, Hager Yasser Mohamed, Alexander Hitt, Valeriia Lomakina, Henning Jiang, Jennifer Jang
Abstract
Boundary representation (B-rep) is the standard format used by modern CAD systems for parametric 3D models. It turns out, the exact same solid can be represented by different B-reps: for example, two engineers using different operations, a geometry kernel rebuilding the file, and an export setting repartitioning faces will lead to different B-reps even though the underlying solid remains the same. We show that existing B-rep encoders are not robust to variation in the B-rep with the same solid on perturbations applied to standard benchmarks, naturally occurring variations inherent to CAD software, and differences in how designers model the same part via a human dataset we created in FreeCAD. The performance of popular B-rep encoders often collapses catastrophically. We propose the canonical region graph, an input representation whose nodes, features and coordinate frame are derived from the solid itself and show theoretical invariance guarantees on repartitioning and rigid motions. It matches the strongest baseline on standard benchmarks, and is stable under every perturbation we test.