Periodic Reporting for period 3 - MYCLOTH (Sustainable Algorithmic Modeling of Personalized Garments)
Reporting period: 2024-02-01 to 2025-07-31
Clothing is a fundamental and indispensable part of human existence, and dressing people has tremendous societal, economical, and environmental impact. The digital technological revolution has not yet achieved its full potential in the garment manufacturing and retail domain, where many of the processes remain essentially unchanged for a century. In addition, the “fast fashion” approach of selling large volumes of garments at very low prices and changing collections very frequently leads to significant waste of natural and human resources: according to recent research, nearly 50% of all manufactured garments are imminently destined for landfill or incineration.
A fundamental, radical change is required to the way clothes are designed, manufactured and delivered to consumers, generating value and attachment via quality instead of quantity and curbing toxic overproduction. The goal of this project is to bring transformative technological advances in geometric modeling and optimization of personalized, custom-fitted and fabricable garments in order to crucially support this change. To reach our goals, we must break away from traditional shape representations and modeling pipelines and develop a dedicated mathematical and algorithmic foundation for digital cloth and garment modeling. Our envisioned theoretical basis of the digital garment shape space will on the one hand facilitate a novel interactive modeling framework to support apparel designers in the creative task of template garment design in a reality-faithful manner, and on the other hand serve as the enabling foundation for automated, algorithmic garment personalization to perfectly fit any human body model. In stark contrast to current practice of standardized confection sizes, our framework will enable on-demand fabrication of custom-tailored clothing while being inclusive of the full diversity range of human shapes, and anticipate the proliferation of digital garment fabrication technology.
A fundamental, radical change is required to the way clothes are designed, manufactured and delivered to consumers, generating value and attachment via quality instead of quantity and curbing toxic overproduction. The goal of this project is to bring transformative technological advances in geometric modeling and optimization of personalized, custom-fitted and fabricable garments in order to crucially support this change. To reach our goals, we must break away from traditional shape representations and modeling pipelines and develop a dedicated mathematical and algorithmic foundation for digital cloth and garment modeling. Our envisioned theoretical basis of the digital garment shape space will on the one hand facilitate a novel interactive modeling framework to support apparel designers in the creative task of template garment design in a reality-faithful manner, and on the other hand serve as the enabling foundation for automated, algorithmic garment personalization to perfectly fit any human body model. In stark contrast to current practice of standardized confection sizes, our framework will enable on-demand fabrication of custom-tailored clothing while being inclusive of the full diversity range of human shapes, and anticipate the proliferation of digital garment fabrication technology.
The MYCLOTH team has designed a new mathematical and algorithmic formulation of digital textile modeling, as well as a novel algorithm for garment modeling directly in 3D based on a 3D scan of the intended wearer, which can be obtained by simple and relatively inexpensive means, e.g. using a tablet equipped with a depth camera. The approach is suitable for professional designers and lay users alike and takes common body poses into account when calculating the garment shape. The result is a customized 3D model of a garment that specifically fits the person it was designed for and accommodates the person's movements and body shape specifics. The garment created with our system does not follow any pre-defined templates and standardized sizes, which means that it is possible to model clothes for a very wide range of people, including persons with missing limbs, scoliosis, dwarfism and other features. This work has been published as a scientific paper in the Computer Graphics Forum journal and open-source code.
Having a custom-fit 3D model of a garment, the MYCLOTH team set out to solve the next posed challenge: producing the necessary data to fabricate the garment physically. We have developed a method for computational patternmaking based on a 3D model of a garment that respects various fabrication constraints, such as reflectional symmetry of seams, enabling the computer to produce a sewing pattern ready for manufacturing. This work has been published at the SIGGRAPH conference and as open-source code.
To enable the achievements of the milestones above, we have developed a number of algorithms for digital geometry processing and human body modeling that constitute standalone contributions. We have developed methods for shape matching and mapping between 3D shapes, which we require to be able to register different poses of a human body. This is a fundamental tool of geometry processing, and in particular, such registration and mapping are necessary for the modeling of garments that fit the wearer in various poses and motions. We have also developed novel methods for human pose and motion modeling, taking advantage of modern machine learning techniques. Understanding the dynamics of the human body shape helps us create algorithms for garment modeling that accommodate these dynamics. Finally, we have been studying the geometry of surfaces and fabrication of 3D surfaces out of 2D sheets by means of developable geometry, which is a fundamental component of clothmaking.
Having a custom-fit 3D model of a garment, the MYCLOTH team set out to solve the next posed challenge: producing the necessary data to fabricate the garment physically. We have developed a method for computational patternmaking based on a 3D model of a garment that respects various fabrication constraints, such as reflectional symmetry of seams, enabling the computer to produce a sewing pattern ready for manufacturing. This work has been published at the SIGGRAPH conference and as open-source code.
To enable the achievements of the milestones above, we have developed a number of algorithms for digital geometry processing and human body modeling that constitute standalone contributions. We have developed methods for shape matching and mapping between 3D shapes, which we require to be able to register different poses of a human body. This is a fundamental tool of geometry processing, and in particular, such registration and mapping are necessary for the modeling of garments that fit the wearer in various poses and motions. We have also developed novel methods for human pose and motion modeling, taking advantage of modern machine learning techniques. Understanding the dynamics of the human body shape helps us create algorithms for garment modeling that accommodate these dynamics. Finally, we have been studying the geometry of surfaces and fabrication of 3D surfaces out of 2D sheets by means of developable geometry, which is a fundamental component of clothmaking.
By the end of the project, we expect to produce a CAD-like user-friendly modeling system for custom made garments that is based on a novel digital cloth model, where users can manipulate virtual cloth as if it were any generic CAD surface. We intend to disseminate this result in the form of open-source software.