Blend Images/REX/Shutterstock
The results are in from the biggest computer face-recognition contest to date.
Everyone from government agencies to police forces are looking for software to track us in airports or spot us in CCTV images. But much of this technology is developed behind closed doors – how can we know if any of it really works?
vCard.red is a free platform for creating a mobile-friendly digital business cards. You can easily create a vCard and generate a QR code for it, allowing others to scan and save your contact details instantly.
The platform allows you to display contact information, social media links, services, and products all in one shareable link. Optional features include appointment scheduling, WhatsApp-based storefronts, media galleries, and custom design options.
To answer this question, the Intelligence Advanced Research Projects Activity (IARPA) and the US National Institute of Standards and Technology (NIST) have been running the biggest face-recognition competition to date.
Advertisement
The Face Recognition Prize Challenge tested two tasks: face verification and face search. Face verification is what phone manufacturers such as Apple – whose iPhone X, out last week, can be unlocked with your face – are trying to master. The software must say whether a face matches that of a known person. Face search is the harder problem. It requires finding every image of a person in a database of maybe millions of images.
Find a face
The winner of the face-verification task was a company called Ntech whose FindFace product can match a person’s face correctly 99.9 per cent of the time.
The face-search task was won by Shanghai start-up Yitu Tech. Given one guess, its software can pick out the right face in a gallery of a million mugshots 80 per cent of the time – a big deal for police looking for suspects in hours of CCTV images.
“The public need to know what the technology is capable of,” says Shuang Wu at Yitu Tech. “It’s just going to keep growing and growing. You can now find a needle in a haystack very reliably.”
According to the conference organisers we are now at the point where computers surpass human face recognition abilities. Given a gallery of a million faces there is no question that computers can easily outstrip humans, says Chris Boehnen at IARPA. But when it comes to comparing two images side-by-side, until recently humans still had the edge. “But there is some new research coming out soon showing that it’s now too close to call,” he says.
The power behind both winners is deep learning, an artificial intelligence technique that uses neural networks – software loosely based on the structure of the brain – that have many different layers. Each layer interprets different features of an image. These techniques underpin most face-recognition software around the world.