Yelp for food, Feedly for news, WeChat for communication, TikTok for entertainment - there are now over 2 million apps to choose from that have become an integral part of our lives. We spend countless hours on our favorite apps, why not customize the experience?
While most mobile applications have a one-size-fits-all graphical user interface (GUI), research shows that many users actually prefer variety in their GUIs. For example, Candy Crush adapts to different styles at different stages of the game, and Line Messenger allows users to change the start screen, friend list, and chat screen. However, in most cases this flexibility is limited and user customizations may be lost when the developer changes the application.
To enable content creators and end users to seriously redesign the interfaces of their applications while maintaining the clarity of content details critical to their usability, researchers at Stanford University propose ImagineNet, a novel and powerful interface Customize new tools.
ImageNet uses neural style transfer models to enable users to apply styles such as artwork to change the visual appearance of mobile applications and their assets - imagine Abstract Expressionist home screens, Brutalist pizza delivery apps, Cubist Pac-Man video games .
"We envision a future where users will expect to see beautiful design in every app they use, and enjoy diverse designs as much as is fashionable today," the researchers write in their paper ImagineNet: Using Neural Style Transfer to Recreate Design application writes.
Style transfer is a computer vision task that extracts styles from reference images and applies them to input images. The Stanford paper proposes a neural solution by adding a new structural component to the original style transfer modeling, which is computed as a non-central cross-covariance between features from different layers of a CNN. The new structure links style elements selected for each location across levels, better transferring styles to interface assets like buttons, frames, etc.
Failing to spot any previous attempts to apply style transfer to GUIs, the researchers compared ImagineNet with style transfer algorithms for realistic images. However, when they tested these style transfer algorithms on the MemoryGame Android app GUI, they found that using these techniques to style the game made it no longer playable.
By minimizing the squared error in the structure between the style and the output image, ImagineNet is able to preserve the details of the GUI while consistently transferring the colors and textures of the chosen art style.
ImagineNet aims to redesign basically all types of applications, allowing users to unleash their creativity and personalize their favorite games.
The paper ImagineNet: Redesigning Applications Using Neural Style Transfer is published on arXiv.