Personalized Image Generation by Exploring Latent Spaces with Swipe Interactions

Abstract

GANs have been shown to generate more realistic images than conventional generative models. By manipulating the latent variables of GANs, users can generate their preferred images. However, GANs' latent space is often high-dimensional and can be challenging to explore. Existing research generates preferred images by asking users to adjust multiple sliders and edit images. In this study, we propose a more lightweight method to generate preferred images by using a simple swipe operation. The results of our experiments demonstrate the potential of the proposed method in generating preferred images.