COMPARING THE PERFORMANCE OF DIFFERENT GENERATIVE ADVERSARIAL NETWORKS ON REALISTIC AND ART IMAGES

Abstract

This research explores the performance of different Generative Adversarial Networks (GANs) on realistic and art images. Specifically, implementation and comparison of outcomes for Deep Convolutional GANs (DCGANs) and Conditional GANs and lastly this study evolves to include Info GANs. The methodology involves training these models on a diverse dataset comprising realistic and art images and evaluating their performance through various metrics. This study aims to comprehensively review GAN design for spatial imaging. By analyzing DCGANs, Conditional GANs, and Info GANs, the research aims to reveal their strengths and limitations in image generation tasks. Through rigorous analysis and comparison, we can look at the capabilities and potential applications of these GAN architectures, contributing to the development of fertility modeling techniques and their real-world implications