Impressive: New Nvidia AI Reconstructs Partially Erased Photographs

April 25, 2018


This is a video demonstration of a new deep learning artificial intelligence program developed by US tech giant Nvidia (best known for their GeForce line of graphics processing units) that can reconstruct digital photos that have had irregular pieces erased in a matter of moments, and with impressive accuracy. Are you thinking what I'm thinking? "All those Instagram photos with the boobs scribbled out!" A simple 'no' would have sufficed.

According to the researchers, they're the first to successfully train a neural network to process irregular shaped holes in images.

Previous deep learning approaches have focused on rectangular regions located around the center of the image, and often rely on expensive post-processing. The goal of this work is to propose a model for image inpainting that operates robustly on irregular hole patterns, and produces semantically meaningful predictions that incorporate smoothly with the rest of the image without the need for any additional post-processing or blending operation.

Think of the possibilities. "Instagram boobs!" Enough with the Instagram boobs already, this isn't the back of the school bus. Wait -- is it? Who's selling candy? I dare you to open the emergency exit! Ooh, ooh, wait -- let's get this tractor trailer to blow his horn! Man, those were the days, weren't they? "I've had some pretty good days since." Lucky you.

Keep going for the video.

Thanks to n0nentity and Lydia, who agrees the key to reconstructing a photograph with a piece missing is digging through the trash to find that piece.

Previous Post
Next Post