Canvases with landscapes, which at one time inspired artists to create true masterpieces, with the help of a new algorithm of artificial intelligence, it became possible to turn them into realistic photographs.
Scientists were able to teach AI to "translate" an image from one "language" to another, for example, from painting to photography and vice versa. Also, manipulations for transforming various objects became possible. For example, with fruits (turning apples into oranges) or with horses (from an ordinary horse to a zebra).
Researchers at the University of California, Berkeley demonstrated the capabilities of the AI algorithm they developed using the example of several paintings by Claude Monet, translating them into the "language" of color photography.
Thanks to the same algorithm, you can do "all the way around", that is, an ordinary photo can be turned into a picture painted in the style of Van Gogh or Cezanne.
The main feature of the new technology is the unpaired data system. Instead of training a machine on the difference between the elements of two paired photographs, scientists have taught artificial intelligence to find the desired match on its own.