PolycubeNet: AI Generates Hexahedral Meshes from Point Clouds
An end-to-end framework for generating polycubes using conditional diffusion models has been unveiled by PolycubeNet. This innovative method takes an input geometry represented as a point cloud and generates a corresponding polycube point cloud directly, eliminating the need for explicit surface segmentation or pre-established templates. At the heart of this development is a dual-latent conditional diffusion architecture that restricts costly self-attention processes to a low-dimensional latent space with fixed capacity. This technique effectively tackles the difficulties associated with automatic hexahedral mesh generation for complex CAD geometries, which often face challenges with traditional polycube methods that depend on complex segmentation and local heuristics that can lead to artifacts or failures with intricate shapes. The research is available on arXiv under ID 2605.20274.
Key facts
- PolycubeNet uses conditional diffusion models for polycube generation.
- Input is a point cloud; output is a polycube point cloud.
- Eliminates need for explicit surface segmentation or predefined templates.
- Dual-latent architecture reduces computational cost of self-attention.
- Addresses challenges in hexahedral mesh generation for complex CAD geometries.
- Published on arXiv with ID 2605.20274.
- Polycube-based hexahedral meshing is known for regular, parameterization-friendly structure.
- Existing methods can produce artifacts or fail on difficult shapes.
Entities
Institutions
- arXiv