By any measure, a score of 99.15% is pretty darn impressive.
That near-perfect record belongs to a new facial recognition system from computer scientists at Hong Kong University.
Needless to say, they’re touting it as the most accurate software of its kind anywhere in the world.
And such pinpoint accuracy provides more validation for the fast-growing field of biometrics – and specifically, for the facial recognition area…
Big Brother’s Best Mate
In an era where Big Brother surveillance is never far away, facial recognition technology has it skeptics, of course.
But there’s no denying that it’s critical in fighting crime and identifying potential threats to society. Particularly, as the global population continues to grow, and spotting such threats becomes more difficult.
That’s precisely the motivation for the electronic engineering team at Hong Kong University, which has created a new computer program that weeds out danger like never before.
You see, humans recognize faces with 97.53% accuracy.
But Professor Xiaoou Tang says, “A couple of months ago, we got a result that actually surpassed the human performance in recognizing faces.”
A 99.15% success rate, to be exact.
Not only does that beat humans’ ability, it also defeats other facial recognition software programs.
It wasn’t easy, either…
Nowhere to Run
Tang and his team put their technology through a range of tests to see if it could identify the same person in photographs, even when the camera angles, lighting, makeup, and other factors were changed.
It passed with flying colors.
As you’d expect, such an overwhelmingly positive result should alert law enforcement agencies and security services to the potential here – particularly when it comes to picking out faces in big crowds.
By snapping photographs and running them through the software algorithm to identify known troublemakers and interpret facial expressions, it could both prevent disturbances before they happen, or nail criminals after the fact.
As Assistant Professor, Xiaogang Wang, says, “We try to develop a computation model to detect and analyze the behavior of the crowd and predict potential crowd disasters. In the traditional way – with video surveillance – they only focus on a small number of objects in a very simple environment. But we target thousands of objects in very complex environments. This problem is very challenging, and we are working in that direction.”
And apparently succeeding, too.
What’s more, the developers believe the potentially lifesaving technology will improve even further, as computer speeds increase.
In the meantime, I guess they’ll have to settle for “only” 99% accuracy.