BeWear
Bespoke AI Wear
Creating standard fashion sizes was a good invention some 150 years ago (E. Butterick, 1863), at the beginning of the industrial age, but has reached its limits today. First, every country and every brand has different size systems (volatile, diverse and by no means universally valid). Second, standard sizes cannot be scaled arbitrarily. Third, pattern shapes are scaled according to certain principles. However, these two-dimensional measurements do not take into account the curvatures of the human body. For example, while three persons may have the same punctual body measurements (i.e. of straight lines with normal tape measure), these can vary significantly in the body's 3D structure. This explains some of the problems of adapting traditional pattern-making techniques.
The business model of BeWear focuses on the automated production of individual patterns based on 3D body scans, created and uploaded by customers from their smartphone to the BeWear platform. BeWear ensures rapid market penetration through sales and distribution in established online shops specializing in the respective segments. In principle, mass customization is relevant for the entire clothing industry. Initially, however, BeWear will focus on those niches with the largest unresolved problems. Besides lifestyle issues (sustainability and individualization), this mainly concerns target groups with fitting problems. BeWear will be the only online portal in Switzerland and Europe, either through e-commerce distribution (back-end implementation of the ASP method) or directly via the BeWear platform, to allow customers to select clothing items from a visual database, to customize these and to have them produced locally at attractive prices.
BeWear solves these problems through local on-demand service provisioning. This is made possible by the ASP method, developed during the author's PhD research. ASP is a completely new, interdisciplinary approach to 3D pattern creation. It enables mobile-phone-based 3D body scanning. Body scan data of moderate to poor quality are sufficient (Ballester et al., 2015). Scans are then mapped to a high-quality digital 3D model, to create an anonymized avatar. This in turn enables processing poor mesh and adding semantic information (e.g. clusters and ganglia).
