A team of researchers deep in the halls of Microsoft has developed a new AI-based application. This software allows users to change photographs into caricatures. The Microsoft Research team that developed the app is comprised of Redmond’s Lu Yuan alongside Stanford graduate Kaidi Cao and Jing Liao from the City University of Hong Kong. So yes, file this project under fun. To create the required results, the team developed an AI that can automatically convert images into caricatures. To train the AI to function as planned, the Microsoft Research team used over 1,000 images to convert. Generative Adversarial Networks (GANs) technology was used to generate caricatures. The first GAN measures the face geometry to create a base model of what the face looks like. Secondly, the GAN applies a texturized artistic layer over the model to make it look as if drawn by hand.
Introduction
As you can see in the image above, the results are improved through each layer. While it is hardly a stunning result, it is a fun way to show the power of basic AI learning. To know more, check out the Microsoft Research paper. Furthermore, the project will be showcased at SIGGRAPH Asia in Tokyo next month. We suspect once the official project is launched, an online app will appear to showcase the AI.
