SkinCells: Sparse Skinning using Voronoi Cells

How to read this

Category

Automatic, sparsity-controlled skinning weight optimization via a continuous weight field, aimed at production and mobile LOD pipelines

Contributions

• Introduces SkinCells, a parameterized family of Voronoi-cell-inspired implicit weight field functions optimized in canonical A- or T-pose rather than on discrete vertices, so a single optimization pass applies across all LOD levels of an asset
• Provides direct, controllable sparsity enforcement, limiting bone influences per vertex through a dedicated sparsity loss, addressing what the authors describe as an open problem in prior falloff-based approaches that offered no numerical guarantee
• Combines the sparsity loss with a Delta-Mush-inspired smoothness objective and a spring-based location objective, evaluated on nine characters with linear blend skinning and five skirt designs, reporting more robust results than standard biharmonic weighting and up to 11x smaller storage than per-vertex weight and index lists

Context

The method continues the classic skinning-weight lineage from heat diffusion (Pinocchio) through bone glow, quasi-harmonic, and biharmonic weighting, and specifically builds on the idea that deformation artifacts under sampled poses should be penalized directly during optimization, an idea the paper credits to Kavan and Sorkine and to Thiery and Eisemann, rather than only enforcing weight smoothness as an intrinsic property.

Correctness

Supported by comparative evaluation against biharmonic and other baselines across characters and cloth, with concrete storage figures. It is an industrial paper from Meta Reality Labs targeting production and mobile use cases, so the benchmarks are practically grounded, but the abstract gives no indication of independent third-party validation of the robustness claims.

Clarity

A long, dense ACM-format paper with an extensive related-work section spanning skinning, Voronoi fields, and physics-based animation. Clear for a technical artist or engineer with a skinning background, though the Voronoi weight-field formulation itself needs careful reading in the method section.

How to read it

First pass: abstract and Figure 1's teaser of characters and skirts. Second pass: the introduction's contribution bullets and the related-work section for how it differs from biharmonic weighting and prior sparsity approaches. Third pass: Section 3's Voronoi weight-field formulation and the three loss terms, plus the results tables covering LOD and sparsity, before adopting the technique in a pipeline.