CHS: Small: Pattern Understanding and Computational Modeling for Textiles

University Of Washington, Seattle WA

Investigators

Abstract

Textile production is a rich domain of fabrication, influenced by generations of creativity. However, the lack of digital design tools creates a gap between the community's creativity and modern manufacturing techniques. The opportunity to learn from massive online repositories of patterns shared by creators could provide access to a novel source of creative ideation. This project will develop new techniques to understand and encode hand-produced patterns in all their complexity, and support modifying them in ways that are machine- or hand-producible. This in turn could help to support customization and personalization of patterns to meet the needs of specific contexts from soft robotics to accessibility to patterns that fit all body shapes. The database of digitized designs developed during this work will serve as a resource for future work empirically analyzing the space of designs that online communities have developed, and support applying deep learning techniques in new design tools. This research will establish the domain-specific languages and tools necessary to represent, learn from, manipulate, verify, and automatically fabricate hand-produced textile patterns and support the eventual goal of digitizing directly from this rich set of online data. The first task is to define a high-level domain specific language as well as a lower-level graph-based model that represents the connections between stitches and components of a garment, a parser and compiler for translating between well formed patterns and this graph representation, and tools to verify pattern correctness for machine and/or hand production. The second task is to develop a hybrid human/machine method for parsing the plethora of patterns currently available on online databases. Visualization tools will be developed based on physical simulation at the strand level to allow users to understand, modify, and verify their patterns. Finally, this research will explore new generative design techniques to support modification of pattern texture and shape as well as construction of novel shapes or additions of shaped elements to existing garments. These will be accomplished by developing optimization techniques with objectives on the gross shape and fine texture, over the discrete space of stitches, and under manufacturability constraints. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.