On the Use of Generative Adversarial Networks for User Interface Designs /

Abstract

The rise of the Internet has led to an increase in the availability of the world wide web and smartphones. About 66.6% of the world’s population has access to both online and offline mobile applications. The success of any digital brand depends on the retention of users on its application. The user interface design plays an essential role in attracting and retaining user’s attention. With the emergence of Generative Adversarial Networks (GANs), researchers have given considerable attention to generative modeling in many areas including image generation, text to image synthesis, and 3D object generation. In this research, an effort has been made to use the power of GANs for mobile application user interface (UI) designs following the layout and a proper color scheme from the distribution of designs in the semantic annotations of mobile UIs in the Rico dataset. This study helps to identify the suitable baseline variation of GAN for a dataset having fewer pixel variations by comparing the most relevant architectures. After training the networks, we explored the insights of the generator model and the discriminator model. Synthetic UIs were studied for design manipulation and the design transformation using latent space exploration. The comparative analysis of the networks has been performed both subjectively and empirically. A UI designer Generator is then separated from the network for mobile UI design suggestions.