August 17, 2022
Digital Yarn for Real Socks
Visual computing scientists at ISTA transfer yarn-based cloth simulation to industrial application
The Wojtan group at ISTA contributes to the fact that Pixar and Disney characters are not naked. Their visual computing of knitted yarn-cloths captures the complexity of garments. Now, they applied their knowledge to a real-world setting that will benefit the textile industry in realizing new fabrics. They present their findings at the esteemed SIGGRAPH conference and by this may contribute to the fact that future humans will not be naked.
Compare animated movies today and twenty years ago. No doubt, you recognize the disruptive advances that have happened in the field of visual computing. This is the hard work of computer scientists like doctoral student Georg Sperl and his supervisor Professor Chris Wojtan at the Institute of Science and Technology Austria (ISTA). However, not only animation artists travel to the annual SIGGRAPH conference to listen to the newest insights from computer scientists. Also technology and industry leaders are scouting. They search for algorithms to improve their production line. “For this project, we applied our efficient and accurate visualization algorithms to real-world problems and based the investigation on real data from the textile industry,” summarizes Sperl the collaboration with the Spanish company SEDDI and the US firm Under Armour.
The beautiful complexity of yarn cloths. It requires an efficient algorithm like the one developed by the researchers to capture the dynamics of thousands of yarn meshes in real-time.
Foundational research finds its use case
What makes Sperls’ simulations of knitted fabrics special is their yarn-based approach. Instead of using a mesh, which only reproduces overall properties of the material, he considers each yarn and its physics. This offers higher control and also captures more of the beautiful complexity of a moving knitted pullover. Yet, the clever algorithm maintains a reasonable computational cost, which means it is still efficient. So far, nobody has ever applied a yarn-based simulation to real industrial data. “We were curious, whether it worked. Real data is tricky. Many parameters are unknown. But the results show, it´s possible and opens up many advantages,” says Sperl and offers an example:
Imagine a textile company wants to add a new fabric to its portfolio, but it does not know the properties of it, how it twists, moves and stretches. Note that the knitting pattern changes the behavior of the cloth significantly and in a complex way. Now, the firm could provide data of different knitting patterns produced from the same yarn. With the novel method, they can then calculate a yarn model, which not only captures the dynamics of the samples, but also numerous other patterns with that yarn. Instead of producing all possibilities and checking them, they could simulate the properties in advance. Such virtual examination would save resources.
“The textile industry is huge and simulation-based approaches are just picking up speed. For us, it is very exciting to shape the methods that might soon be implemented everywhere around the world.”
Georg Sperl, Rosa M. Sánchez-Banderas, Manwen Li, Chris Wojtan, and Miguel A. Otaduy. 2022. Estimation of Yarn-Level Simulation Models for Production Fabrics. ACM Transactions on Graphics, Vol. 41, No. 4, Art. 65 (Proceedings of SIGGRAPH). DOI: 10.1145/3528223.3530167
The research project was funded in part by the European Research Council (ERC) Consolidator Grant 772738 TouchDesign.