None of these people exist: portraits created by artificial intelligence

Computer technology is developing at a cosmic speed, and you and I are practically living in that “future”, which we watched a fantastic movie about in our youth, managing only to pick our jaw off the floor. And here is another proof of this.

The Generative adversarial network is a state-of-the-art smart network that is just over four years old. It was designed to create realistic human portraits based on data collected from real photographs. This means that the GAN creates a photo of the face of a person who does not actually exist.

For fun, it's worth seeing what the capabilities of this artificial intelligence were right after its development five years ago:

Agree, at first he did not cope very well and one would have thought that the idea was hopeless. Well, now look at the results displayed by the neural network now. The thought that not one of these people is in reality becomes even uneasy.

This is how the process looks in progress on video.

Here are some more portraits of "virtual" people.

The creators of AI went further and "taught" their brainchild to create not only human faces, but also the interiors of rooms, animals and even cars.

Watch the video: A Style-Based Generator Architecture for Generative Adversarial Networks (April 2024).