18106 Nis, Serbia +381 62 9648511

Deep Learning at NVIDIA: AI Reconstructs Photos with Realistic Results (Image Inpainting for Irregular Holes Using Partial Convolutions)

Researchers from NVIDIA, led by Guilin Liu, introduced a state-of-the-art deep learning method that can edit images or reconstruct a corrupted image, one that has holes or is missing pixels. The method can also be used to edit images by removing content and filling in the resulting holes.
It will only get better once it has been trained well enough (and we know that nVidia doesn’t suffer from lack of GPUs!), and we are waiting for updates on potential release of this software as a stand-alone tool or a plugin for Gimp or some similar photo manipulation software.

As one youtube user in the comment section put it, this tool really is a photoshop jujitsu.

Here are some more gems from the comment section:

Learn more about their research paper “Image Inpainting for Irregular Holes Using Partial Convolutions”: https://nvda.ws/2K1G243

Related articles

Scientists Construct Bio-computer Made From Living Human Cells

We would benefit greatly from surpassing the limitations of our monkey brains. If we are to construct such AI, we need to make sure we can understand it, and hopefully moderate it – ahem – […]

Learn More

The Poetry of Science: Richard Dawkins and Neil deGrasse Tyson

It really is a pleasure to see two people with clear minds and sound logic have a constructive conversation: no logical fallacies, just a normal conversation on important topics – “a kind of mutual tutorial […]

Learn More

What if NASA had the US Military’s Budget? ($600B)

“Space, the final frontier. Except, we don’t have cool spaceships because the US government has a bit of a thing for aircraft carriers” Once I realized that the US military’s budget is so much thicker […]

Learn More

We would love to hear what you have to say about this: