Watch: Nvidia’s GuaGAN Augments Jottings into Photorealistic Art

By: Norhan Mahmoud 

CAIRO, Mar. 27 (SEE)- To the left it’s a set of childish scribbles. To the right it’s almost oil on canvas. This is what a screen displays as users experience Nvidia’s GauGAN transforming their ideations into works of art. 

What is GauGAN?

“It’s like a coloring book picture that describes where a tree is, where the sun is, where the sky is,” said Bryan Catanzaro, VP of Nvidia’s Applied Deep Learning Research. 

Here’s how it works; users pick an object from the bottom list, then segmentation maps are formed. 

And, once tagged- as rock, water, sand, etc- “the neural network is able to fill in all of the detail and texture, and the reflections, shadows, and colors.”

https://youtu.be/p5U4NgVGAwg

How did GuaGAN get its name?

“Gua” is a special nod to France’s post-impressionist Artist Paul Gauguin. While, “GAN” is the abbreviation of new generative adversarial network, an unsupervised machine learning model. 

Interestingly, no two users will get the same output even if their doodles are atypical. There are no copycats with GuaGAN, as the system is loaded with about one million images which the GAN utilizes to generate enthralling sceneries. 

“The real advance is the ability to synthesize images with a lot more persity and more fidelity than in the past,” said Catanzaro.

Catanzaro believes that this tool is ideal for people making virtual worlds like architects and designers, “to train robots and self-driving cars.”

“I really think this technology is going to be great for the dreamers of the world.”