NVIDIA has launched a new application with which you can get images from realistic landscapes even if you have very little artistic skills. This tool can convert doodles and sketches into photorealistic landscape images in real time thanks to artificial intelligence.
It is now available for download in a free beta form, though You can only use it if your computer is equipped with an NVIDIA RTX GPU (a GeForce, Quadro or Titan graphics) and must be based on the Windows 10 operating system.
Canvas is based on the GauGAN AI paint tool, which NVIDIA Research developed and trained using 5 million images. When the company first introduced GauGAN, NVIDIA Vice President Bryan Catanzaro, described his technology as a "smart brush".
Available materials and how to change a drawing into a large image
Enough with draw lines and shapes on the virtual canvas, and the application will automatically convert each brushstroke into the material you want (you can choose things like whether a drawing is grass, rocks or clouds). The beta version of the application has 15 "materials" available to convert a sketch into a realistic image.
This tool is developed with PyTorch, a deep learning feature. The test carried out by the Xataka colleagues demonstrates the great capacity of GANs to produce very good results.
For this the system has been fed with five million photos that allow the discriminator —One of the key elements of the process— filters so that synthetically created pixels make sense.
This is not Nvidia's first notable foray into the world of artificial intelligence. Let us remember his unique tribute to the title of Bandai Namco, for which he created an artificial intelligence system that simply playing 50,000 games of Pac-Man has recreated the game from scratch respecting the mechanics and rules of the game.