Deep Learning Based Apparel Product Development System

Abstract

The apparel industry is one of the biggest, yet growing areas of business in the world. The objective of this research was to implement a solution that reduces the difficulties faced by the apparel industry when producing garment items in an efficient and timely manner. With the use of Generative Adversarial Networks (GANs) and Regional Convolutional Neural Networks (RCNNs), the expectation is to generate brand new, unprecedented garment items using existing garment items and to identify the basic pattern blocks of generated garment images with high accuracy. Through the experimentation and analysis, we were able to generate new garment images by employing the GAN with an acceptable level of accuracy and was able to identify the basic blocks of the garments with high accuracy Instance Segmentation. Hence, this provides a unique solution that combines both fashion designer’s and pattern maker’s expertise areas at once, which could serve as a perfect platform in optimizing the product development process in the apparel industry.