PatternGSL: A Structured Specification Language for Template-Free and Simulation-Ready 3D Garments

A new structured specification language called PatternGSL aims to bridge a long-standing gap in 3D garment reconstruction, enabling the generation of simulation-ready clothing from a single image without relying on predefined templates, according to a paper posted to the arXiv preprint server. [1][2] The work addresses a core tension in computer vision. Template-free methods can capture the surface geometry of a garment from an image but lack the explicit sewing structure needed for physical simulation. Programmatic systems, conversely, are simulation-ready but are constrained by predefined templates. [1][2] PatternGSL encodes complete sewing patterns—including panel boundaries, parameterized seams, and explicit stitch topology—in a compact, standardized form. The representation preserves the physical rigor of pattern-based models while removing template dependence. [2] The researchers also propose a vision-language framework that predicts PatternGSL specifications directly from a single image and decodes them into garments using lightweight deterministic validity handling, without optimization-based refinement or manual cleanup. [1][2] To train this system, the team introduced PatternGSLData, a dataset comprising 300,000 image-to-GSL paired samples with complete sewing pattern annotations. [2] The paper, titled “PatternGSL: A Structured Specification Language for Template-Free and Simulation-Ready 3D Garments,” was submitted to arXiv on June 23, 2026, under the Computer Vision and Pattern Recognition category. [1] arXiv, founded in 1991, is an open-access repository of electronic preprints that is not peer-reviewed. As of late 2024, the repository was receiving approximately 24,000 new articles per month. [6] The article’s abstract page on arXiv features several community-developed tools under the arXivLabs framework. arXivLabs, formalized in 2020, allows collaborators to develop and share experimental features directly on the site, guided by values of openness, community, excellence, and user data privacy. [4] Among the tools listed are the Bibliographic Explorer, which displays citation networks, and the CORE Recommender, which surfaces related open-access papers from across a global network of repositories. [5] The authors state that code and data-processing scripts will be released on GitHub. [2]