A novel approach in predicting virtual garment fitting sizes with psychographic characteristics and 3D body measurements using artificial neural network and visualizing fitted bodies using generative adversarial network

Advances in technology have brought accessibility to garment product fitting procedures with a virtual fitting environment and, in due course, improved the supply chain socially, economically, and environmentally. 3D body measurements, garment sizes, and ease allowance are the necessary factors to ensure end-user satisfaction in the apparel industry. However, designers find it challenging to recognize customers’ motivations and emotions towards their preferred fit and define ease allowances in the virtual environment. This study investigates the variations of ease preferences for apparel sizes with body dimensions and psychological orientations by developing a virtual garment fitting prediction model. An artificial neural network (ANN) was employed to develop the model. The ANN model was proved to be effective in predicting ease preferences from two major components. A non-linear relationship was modeled among pattern parameters, body dimensions, and psychographic characteristics. Also, to visualize the fitted bodies, a generative adversarial network (GAN) was applied to generate 3D samples with the predicted pattern parameters from the ANN model. This project promotes mass customization using psychographic orientations and provides the perfect fit to the end users. New size-fitting data is generated for improved ease preference charts, and it enhances end-user satisfaction with garment fit.