Design of Customized Garments Towards Sustainable Fashion Using 3D Digital Simulation and Machine Learning-Supported Human–Product Interactions

Abstract

This paper put forward a new interactive design approach for customized garments towards sustainable fashion using machine learning techniques, including radial basis function artificial neural network (RBF ANN), genetic algorithms (GA), probabilistic neural network (PNN), and support vector regression (SVR). First, RBF ANNs were employed to estimate the detailed human body dimensions to fulfill consumers’ ergonomics requirements. Next, the GA-based models were developed to generate the formalized design solutions following the consumer profiles (demands). Afterwards, the evaluation model was established to quantitatively characterize the relations between consumer profiles and garment profiles from the generated design solutions. The design solutions would be digitally demonstrated and recommended to the consumer following the evaluation results in descending order. Meanwhile, the PNN-based models were created to predict garment fitness based on virtual try-on. Moreover, the SVR-based self-adjustment mechanism was built to estimate and control garment design parameters according to the consumer’s feedback. Based on these mathematical models, the approach enhances the interactions among digital garment demonstration, the designer’s professional knowledge and the user’s perception to find out the most relevant design solution. The effectiveness of the new approach was verified by a real application case of leisure pants customization. The results show that the proposed method can powerfully support the designers’ quality personalized design solutions for consumers more accurately, fast, intelligently, and sustainably, compared with the existing approaches. More importantly, it also establishes an effective and reliable communication channel and mechanism among consumers, fashion designers, pattern designer, and garment producer.