Reproduce “Spatially Accelerated Winding Numbers for Curved Geometry” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jacob Spainhour, Brad Whitlock, Kenneth Weiss
The authors extend hierarchical Taylor-based generalized winding number acceleration to NURBS curves and trimmed NURBS surfaces. The method subdivides input, stores geometric moments in a hierarchy, approximates distant clusters, and directly evaluates nearby curved geometry. Across 20,000 SVG shapes, more than half had over 10× query-time speedup against direct evaluation; preprocessing is additional, and accuracy disagreements occur near ambiguous half-integer GWN values.
Key insights
- Directly leveraging parametric NURBS representations within a spatial acceleration framework preserves geometric fidelity and avoids the accuracy compromises inherent in discretizing curved inputs.
- The combination of adaptive subdivision for balanced BVH construction and precomputed moment data is crucial for achieving sub-linear complexity and high accuracy when approximating GWN for curved geometry.
- The 'far-away' criterion (β parameter) and Taylor expansion order provide effective controls for balancing approximation precision against computational runtime during query evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.