NVIDIA AI Transforms Doodles into Photos

NVIDIA AI Transforms Doodles into Photos
NVIDIA has created an artificially intelligent program that can transform drawings into high quality photos.

(CCM) — NVIDIA Research has announced in a blog post GauGAN, its new artificially intelligent software that can transform unsophisticated doodles into lifelike images.

The program, named after the the post-Impressionist French painter Paul Gaugin, uses "generative adversarial networks," or GANs, to convert segmentation maps into photorealistic images. It does so by using its AI "knowledge" of what the real world looks like and matching the doodle to this through inference.

It is also trained on millions of images to fill in the blanks that may be present in a doodle. For example, a user could draw a pond and the software might add tress and rocks as reflections in the water, giving the sketch a boost of realism with ease.

This software could turn out to be especially useful for professions that require sketches during collaborative brainstorming. NVIDIA's blog post cites everyone from architects to urban planners to landscape designers and game developers as potential beneficiaries of such software.

"It’s much easier to brainstorm designs with simple sketches, and this technology is able to convert sketches into highly realistic images," said Bryan Catanzaro, vice president of applied deep learning research at NVIDIA.

Image: © iStock.