Computer Vision / Video Analytics

Inception Spotlight: AI App Can Create A Cartoon Emoji of Yourself

Looking to create a cartoon emoji version of yourself? This new AI app can help. 

A new deep learning-based messenger app can turn users into personalized cartoon characters. 

“Each avatar is unique and personalized to its user,” the company explained. “In this app you can be anyone you want: a human digital copy of yourself, a unicorn, a rabbit or even a tiger!” 

Developed by startup Magic Unicorn, a new member of the Inception Program, NVIDIA’s start-up incubator, the app is the company’s flagship product.

“Our team believes that the future of online communication lies in AI. We’re confident in the app’s cutting edge technologies attracting millions of users and providing them the most entertaining and unique experience of communicating online,” said Ashot Gabrelyanov, CEO of Magic Unicorn

In the back end, Magic Unicorn relies on NVIDIA TITAN V GPUs, with the cuDNN-accelerated TensorFlow deep learning framework to train their models. 

The app relies on a proprietary data set of over 20,000 3D face scans. In May of this year, the company added the ability to generate human-like characters.

The Magic Unicorn team capturing 3D face scans for use in the app. Source: Magic Unicorn

“The process of avatar generation starts by constructing a user’s 3D face model from a selfie shot,” company engineers explained. “With this single photo, one neural network builds a 3D mesh of the user’s head. Then the mesh goes through the automatic stylization process that makes it look more cartoonish. Afterwards, it is used as a basis of the avatar’s head.”

Avatar generation process. Source: Magic Unicorn

“Next, other neural networks swing into action. Face parts such as hair, eyebrows, facial hair and colors are classified and synthesized (generated) onto the mesh in the appropriate style,” the engineers added.

Feature detection example. Source: Magic Unicorn

Once the avatar is generated, facial expressions are captured with a CNN and then reflected onto a user’s 3D face model. The app’s ML models can process up to 60 frames per second and recognize an innumerable quantity of micro-expression combinations, the company said. 

Video example from the app.

The company says their app has an advantage over current avatar generating applications due to their cross-platform application (iOS and Android), as well as their automatic personalization powered by AI.

Video example from the app.

The app is available on both the App Store and Google Play

Read more>

Discuss (0)