NVIDIA's neural network turns primitive images into realistic images (4 photos + video)

Developers around the world are actively promotingthe use of deep learning (a set of machine learning techniques). At the same time, NVIDIA engineers managed to create an original graphical application that allows you to create a photographic image based on a schematic drawing of the user.

& times

The application was named GauGAN, connectingin itself are the abbreviations of the name of the artist Paul Gauguin (Paul Gauguin) and used in the program of generating adversarial networks (GAN), helping to get a realistic photographic image. GauGAN's work imitates a smart brush, which fills certain fragments of a picture with specified realistic photo elements. At the same time, separate parts of the photographic picture created will interact with each other. For example, the environment, plants, sky will be reflected in the water.

GauGAN editor besides the environment andlandscapes can generate in the picture buildings, roads, people, vehicles. Another important feature of GauGAN is the ability to change artistic styles, allowing you to get a picture in accordance with the style of your favorite artist. In addition, the user can already on the finished picture to change the time of day, adding or reducing the illumination of the image, or the time of year. However, while the graphic editor based on AI is available only to developers, when the program becomes available to anyone, it is not known.