Humans are great at spotting anomalies, but terrible at concentrating on what we’re supposed to be keeping an eye on. That’s why inspectors are able to sneak so many guns past TSA security agents. Machines have unlimited attention, but are often less good at detecting things like suspicious packages—in part because they need to know precisely what it is they’re looking for. A new camera from Movidius, however, defies this rule.
The new Hikvision camera uses neural networks to “automatically detect anomalies such as suspicious packages, drivers distracted by mobile devices, and intruders trying to access secure locations.”
Tech-wise, the breakthroughs are the camera’s stereo 3D imaging, and a new chip that’s powerful enough to do all the analysis in real time inside the camera itself, instead of connecting to a bigger computer in the cloud. According to the press release, the new camera managed 99% accuracy in tests, but the nature of those tests isn’t revealed.
Deep learning like this is what allows you to type something like “muffin” into your iPhone’s photo search field and get back not even just pictures of cakes, but specifically all the photos of muffins you’ve ever taken. But it’s also able to track activists and protestors, or pretty much anything else its owners, private or government, want it to track.
It’s both terrifying, and also a great new investigation technique for TV crime dramas.
When used as a security camera, it’s able to not only detect movement but to see what is moving, distinguishing tumbleweed from a car or a human intruder, say. But it has also been used to locate a lost senior citizen using facial recognition, which is an ability with obvious privacy concerns.
It’s ability to spot texting drivers coupled with license-plate reading, and automatic ticketing, could we finally have a solution to careless, law-flouting types on the road?
The University of Manchester now touts itself as the “Home of Graphene.”
As a technology lover you may like our daily important technology news series. It’s a playlist and you can retrieve the total day-wise videos. Please consider subscribing our YouTube channel:
Author: Technology and Beyond

