This neural network takes a bottom-up approach to GPS positioning

Even the best urban explorers get lost occasionally, which is probably how phones have proven themselves as essential tools for navigation in a crunch. In unfamiliar environments, you might even walk a couple of blocks in the wrong direction before realizing you have to turn around—and by then you’re even farther from your destination than when you started.

Researchers at the University of Cambridge have created a capable neural network they call PoseNet. Better than the methods of satellite positioning, PoseNet can deduce your location with a single snapshot in just five milliseconds—60 times faster than a blink. Moreover, the results are just as accurate as any other GPS with a margin of error of only seven and a half feet. The system can even tell which way you’re facing give or take four degrees. The technology is capable of operating indoors, making it a viable alternative for navigating office buildings and subways.

BASIC MEMBER CONTENT
This content is available for Basic Members.
Already a member, log in