3D Design based on Convolutional Neural Network for Customized Clothing

This research presents an innovative 3-D designing methodology that develops personalized clothing in the field of sustainable fashion incorporating deep learning approaches such as Radial Basis Function Convolutional Neural Network (RBF-CNN), probabilistic neural network (PNN), genetic algorithms (GA), and support vector regression (SVR). For the primary phase, a model of human body proportions is formulated using RBF CNNs with the consideration of consumers' requirements so that the final product can be properly fitted. Afterwards, when GA-based models are ultimately used to construct customized design concepts which are in accordance with customer attributes and goals. And, in turn, those suggestions are presented and promoted to consumers through digital means that are built-in with their evaluation score factor. Additionally, PNNs are being developed to be able to evaluate fit using the virtual try-on simulations. In addition, the author creates an SVR-based self-adjusting mechanism that helps to measure and control the fitness feature of cloth design using the feedback of users. The effectiveness of this one-of-a-kind approach is exemplified by a real-world case research of personalization of leisure pants, which was going on recently.