NVIDIA Opens Up The Code To StyleGAN - Create Your Own AI Family Portraits

This week NVIDIA's research engineers open-sourced StyleGAN, the project they've been working in for months as a Style-based generator architecture for Generative Adversarial Networks.

The machine learning technology is for generating new images that mimic the appearance of real images. With StyleGAN, unlike (most?) other generators, different aspects can be customized for changing the outcome of the generated images. StyleGAN is able to yield incredibly life-like human portraits, but the generator can also be used for applying the same machine learning to other animals, automobiles, and even rooms.

These people are not real, they were generated by NVIDIA's newest open-source project.

More examples can be seen in this StyleGAN video by NVIDIA Research.

As of this week, StyleGAN is now open-source for your own experimenting or just to try out this machine learning code yourself. StyleGAN depends upon NVIDIA's CUDA software and GPUs as well as TensorFlow. StyleGAN can work on both Windows and Linux but NVIDIA "strongly recommends Linux for performance and compatibility reasons."