Air Force hires Trueface for facial recognition on bases
Edge computing will improve industrial processes in manufacturing, and enable facial recognition in retail environments and hotels.
The Air Force has signed a deal with Trueface, a developer of computer vision systems, to provide facial recognition, license plate recognition and weapon detection for an unnamed air base.
Both the Air Force and Trueface declined to say which base the technology would be used but Shaun Moore, CEO of Trueface, said the company had more plans to work with the government and military institutions on these kinds of projects. “The goal here is to protect the assets and people on base,” Moore said in an interview.This comes after the Air Force hired the company earlier this year to conduct research on how to use facial recognition software on bases. The new pact was built out of the initial research they worked on this year, according to a Medium post from Mason Allen of Trueface.
“So we’re running a facial recognition scan to ensure the people entering the Air Force base are who they say they are and should be allowed in. We’re also running the license plates of most cars and we’re looking out for weapons in areas where weapons should not be present,” he added.
Trueface has deployed its technology in a variety of ways since starting out in 2013. It was originally a smart access company focusing primarily on building cameras and providing facial recognition software.
But by 2017, executives decided to scrap the camera idea and focus on the facial recognition software.
“We saw a really an opportunity to use the technology for heightened or improved security to reduce the inefficiencies of everyday things like access control. We saw a much larger opportunity to have an impact than providing cameras with the software on it. Our focus became the software,” Moore said.
They now have honed in on three main use cases with escalating degrees of difficulties. The easiest involves software for one-to-one recognition, which you can usually find with things like FaceID for unlocking your phone or for access to bank accounts.
The next level, which Moore called one-to-few, involves facial recognition for airports, customs or buildings and work environments. These are also a relatively low levels of difficulty because as long as they are provided with photos it is easy to identify people walking into a building with an ID card.
The third use case is where there has been some public backlash. One-to-million facial recognition software, meaning ones that can pick people out of a crowd, have been the most hotly debated considering reports of wild inaccuracy and a particular propensity for mistakes with non-White faces.
“The concern around misidentification of people of different ethnicities and even genders and ages is an industrywide problem that we’re solving. It’s a data problem. The reason that it exists today is because the data that the algorithms are trained on are skewed in one direction or another,” he said.
“For us to have perfect accuracy, we need perfect representation in the groups we’re looking at. For the last two or three years now, we’ve seen the cost of data go down, the accessibility of data go down, although it was a problem two years ago, the industry as a whole has been mitigating that.”