Predicting 3d Garment Fit in Digital Product Development: Selecting Input Variables

Three-dimensional (3D) technologies such as 3D digital prototyping have recently been used in apparel engineering to reduce the number of costly physical prototypes. This technology can simulate garment drapes by digitally sewing 2D patterns and placing them on a 3D avatar. Yet, current limitations in simulation accuracy mean physical prototypes are still indispensable. Integrating AI technologies shows promise in accelerating digital transformation in the apparel industry. AI can support decision-making in apparel design, size and fit prediction, and fashion recommendations. It can also assist decision-making in technical design by predicting garment fit before making physical prototypes or offering solutions for misfit areas. Research indicates that models such as Naive Bayes (NB), Support Vector Machine (SVM), and Artificial Neural Networks (ANN) offer robust solutions for fit evaluation. However, they require further testing across diverse garment styles and demographics. Developing a nuanced classification system for fit conditions could improve model accuracy, ultimately supporting the industry's shift towards digital and AI-enhanced product development. In this study, we propose a framework to achieve such a system.