Reproduce “Bayesian Differentiable Physics for Cloth Digitalization” 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

Deshan Gong, Ningtao Mao, He Wang

A Bayesian Differentiable Physics model is introduced to accurately digitalize real cloths by inferring their complex, heterogeneous, and stochastic physical properties from limited, controlled Cusick drape test data. This approach enables faithful digital replicas and realistic garment simulations by capturing fine-grained material behaviors.

Key insights

  • Precisely controlled data, like that from the Cusick drape test, is fundamental for accurately digitalizing specific cloth materials, surpassing the capabilities of models trained on casually captured data.
  • A Bayesian framework within differentiable physics can effectively model and infer the inherent material heterogeneity and stochastic draping dynamics of real cloths, leading to more realistic simulations from limited data.
  • The learned posterior distributions of physical parameters offer a quantitative means to characterize and differentiate cloth types, which is valuable for applications like material science and quality control.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.