Abstract
dc:description.abstractNowadays, cloth simulation is widely applied in 3D entertainment and fashion design. It also has been popular in computer graphics for decades and various simulation methods have been proposed. In both applications and research, cloth simulation realism is an important concern. Among the existing cloth simulation methods, physics-based cloth simulation methods integrate physical laws to explicitly model cloths' physics and govern cloths' mechanical behaviors. Compared with the other existing methods, they can simulate visually more pleasing and mechanically more accurate cloths. Moreover, they are capable of synthesizing various cloths made from different materials by using different cloth physical parameters. However, a physics-based cloth simulator's simulation realism may be restricted by its coarse cloth physical model and inaccurate cloth physical parameters. To pursue higher simulation realism, this research introduces more fine-grained cloth physical models and accurate cloth physical parameter estimation methods. In our first and second works, we propose two differentiable cloth simulators with more fine-grained cloth physical models that can capture cloth yarn-level mechanical behaviors and model cloth material heterogeneity. By using our novel fully-differentiable physical models, they can leverage gradient-based optimization methods to accurately estimate cloth physical parameters and then reproduce the observed cloths. In our third work, we introduce a cloth simulator that can simulate cloth time-dependent persistent wrinkles by integrating our novel time-dependent friction and plastic model. Through exhaustive experiments, we demonstrate our differentiable cloth simulators are accurate in estimating cloth physical parameters and reproducing the observed cloths. We also demonstrate our third simulator can plausibly simulate cloth time-dependent persistent wrinkles. Through these three works, we gain a significant improvement in cloth simulation realism.
Degree
thesis:*- Name dc:type.qualificationname
- Ph.D
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- University of Leeds
- Year dc:date.issued
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Gong, Deshan
- Advisors dc:contributor.advisor
-
- Wang, He
- Bulpitt, Andrew
Rights
- Language dc:language.dcterms_iso639_2
- eng