I briefly noticed a news story on TV a few days ago which featured a piece of facial recognition technology worn as glasses (like Google Glass) which included big claims about being able to identify a face from a database of thousands of persons of interest. I think the news story linked to above is that story, about X6 spy glasses, Osterhout Design Group, Defense Intelligence Agency (USA Govt) and Dr Brian Lovell, a professor at the University of Queensland and CTO at Imagus Pty Ltd. Dr Lovell has a very impressive CV but all the same I was not impressed in 2012 with the way he lightly dismissed the capabilities of human face recognition in an appearance on the Catalyst science TV show from the ABC. I’ll be impressed by the X6 technology when it is used and tested in real life applications. The period of time that human facial recognition capabilities have been used and tested in real life applications is measured in millions of years, so I think the artificial versions of facial recognition might have a bit of catching up to do.
On a similar theme I have found some interesting recent news stories comparing the facial recognition databases of Facebook and the FBI. It looks like Facebook’s DeepFace is superior in important ways compared to the FBI’s Next Generation Identification. I think it is important to consider the value of such databases, as well as human face recognition, as tools in the investigative process rather than as producers of forensic evidence, and for sure these technological facial recognition databases can use (memorize?) huge volumes of images. But regarding the actual process of face recognition, I’m still to be convinced that there is any technology that can do what humans, including human super-recognizers, can do. As Russell Brandom wrote in The Verge “While there are plenty of contractors who are willing to promise “near-human” recognition capabilities, real facial recognition is much harder than the industry lets on.”