Slashdot Mirror
DeepMind Self-training Computer Creates 3D Model From 2D Snapshots (ft.com)
DeepMind, Google's artificial intelligence subsidiary in London, has developed a self-training vision computer that generates 'a full 3D model of a scene from just a handful of 2D snapshots," according to its chief executive. From a report: The system, called the Generative Query Network, can then imagine and render the scene from any angle [Editor's note: the link maybe paywalled; alternative source], said Demis Hassabis. GQN is a general-purpose system with a vast range of potential applications, from robotic vision to virtual reality simulation. Details were published on Thursday in the journal Science. "Remarkably, [the DeepMind scientists] developed a system that relies only on inputs from its own image sensors -- and that learns autonomously and without human supervision," said Matthias Zwicker, a computer scientist at the University of Maryland who was not involved in the research. This is the latest in a series of high-profile DeepMind projects, which are demonstrating a previously unanticipated ability by AI systems to learn by themselves, once their human programmers have set the basic parameters.
I'm not sure why this is a big deal. MS had the tech for this about 5 years (Send us a buncha 2d pictures and we'll turn it into a VR set) extrapolating to models isn't that far of a reach.
I know this because I met this girl on tinder and made a judgement call about how big her boobs were from the pics, but when I got her naked they were WAY HUGER and by a lot. No way a computer could have modeled those monsters from just a photo.
Not only was that anticipated, and not only have computers been "teaching" themselves for years in AI once the basic parameters are set, that is exactly what neural networks were DESIGNED to do nearly 50 years ago when they were invented
"First they came for the slanderers and i said nothing."
that is exactly what neural networks were DESIGNED to do nearly 50 years ago when they were invented
Neural networks were "invented" millions of years ago, pathetic human
Please stop posting links to paywalled web servers.
You should talk to APK he needs your help.
How well does it work on 2D anime drawings?
#DeleteFacebook
The system, called the Generative Query Network, can then imagine and render the scene from any angle [Editor's note: the link maybe paywalled; alternative source],
Why didn't you just use the definitely non-paywalled source?
systemd is Roko's Basilisk.
The technique of solving several images into a 3 dimensional object is called photogrammetry. https://en.wikipedia.org/wiki/Photogrammetry. There are many software solutions that do this, and there are non software solutions that were used to for example create the topographic maps published by the USGS where elevations were extracted from aerial photographs. Most just output 3D geometry information. The new thing might be the way deep mind uses the 3d model results.