Automatic 3D garment modeling by continuous style description

Manually modeling of 3D garment is highly time-consuming and professional expertise demanding, and can only produce limited garments (Fig. 1(a)). The ability to create a diverse set of 3D garments is required with the trend of online fashion and apparel mass customization. This issue has been recently tackled with a fully automatic garment transfer algorithm [Brouet et al. 2012] based on pattern grading. Content creation techniques such as [Xu et al. 2012] introduce set evolution as a means for creative 3D shape modeling. However, current component assembly-based 3D shape modeling were just designed for discrete properties. An important observation is that style of many components garments can be characterized by the ratio of area and boundary length. Thus, we propose a simple but effective garment synthesis method that utilizes such a style description, instead of discretizing the style space. Results show that the method is able to produce various reasonable garments efficiently.