The software uses generative adversarial networks to transform even the worst of sketches into photorealistic landscape.
We all have experimented with the old-school MS Paint and let our imaginations run wild. Now, what if those embarrassing doodles could be transformed into actual work of art?
Nvidia has developed a new AI-driven software that is capable of doing exactly that. Called GauGAN image creator, the software uses generative adversarial networks to transform even the worst of sketches into photorealistic landscape, Engadget reported.
The technology behind the software is a deep learning AI that has been trained on a million images and Nvidia describes it as a ‘smart paintbrush’.
At the core of the tool are three parts; a paint bucket, pen, and pencil. These parts transform the blue line you drew for a waterfall, into an actually detailed, photorealistic waterfall. Similarly, circles drawn can be transformed to beautiful clouds, lumps into rocks, and so on.
The software is currently being only demoed. However, the developers envision it to be useful for everyone including architects, urban planners, even video game developers.