A deep learning-driven clothing style generation and style transfer system

The increased interest in artificial intelligence (AI) being utilized within the realm of fashion design has improved the creation of new systems that generate clothing styles and transfer those clothing styles through deep learning (DL). Many of the existing systems do not balance the creative and traditional aspects of the clothing styles generated, indicating the need for an effectively designed system. This research aims to present a DL- driven method to generate and transfer clothing styles. A Green Anaconda Algorithm-based Style-based Generative Adversarial Network 3 (GAA-StyleGAN3) is suggested to generate and transfer clothing styles. The data represented different types of clothing within their respective fashion ethnicities. The dataset was processed to improve visual clarity and color consistency using bilateral filtering (BF) and histogram equalization (HE). Feature extraction was performed using Vision Transformer (ViT), which allows for a robust multi-scale representation of the textures and outlines of clothing. The GAA-StyleGAN3 system uses StyleGAN3 for generating clothes in high resolution and uses the GAA as an optimizer to optimize hyperparameters. GAA-StyleGAN3, implemented in Python 3.11, produces high levels of realism and style retention across a range of different clothing categories. Experimental results show a generation time of 12.18 s, a total storage consumption of 238.50 MB, or resource consumption of 55.34%, a running time of 0.035 s, and memory consumption of 128.50 MB, along with a satisfaction rating of 9.8 from users and a general rating of 0.79. GAA-StyleGAN3 automates fashion design by allowing users to generate clothing designs.