Scientists at the University of Modena have developed an artificial intelligence that rebuilds an entire image of a person that is partially hidden by any object. The technology opens the door to eliminate photobombs forever — but could also be effectively used for surveillance.
The research paper is titled “Can adversarial networks hallucinate occluded people with a plausible aspect?” And the answer, as the test images show, is a resounding yes.
The results are in fact amazing. The new technology — which uses generative adversarial neural networks to generate these results — could eventually be used on smartphones to eliminate unwanted objects from any photo and have a clean picture of anyone, no matter the environment.
Imagine capturing the perfect moment on a crowded space and have the photo ruined by someone passing between your camera and your subject. This tech will automatically make the photo free of annoying people, birds, or any other kind of object.
But of course it goes way beyond consumer applications. The most obvious and scary one is surveillance. As processor speed increases, technology like this could be applied to a video feed in real time to better track people.
This will allow authorities running this AI on a CCTV network to clearly isolate and follow anyone through New York or London’s busy streets, eliminating any objects or people that get on the way.
It’s only a matter of time that this gets extended to other occluding objects. The most obvious, of course, is clothing. Reality is getting destroyed faster than anybody could have imagined.