How Well Can Algorihms Recognize A Masked Face?; How China Built Facial Recognition For People Wearing Masks
Tom Simonite posted a May 1, 2020 article to the security and technology website, WIRED.com, discussing a subject that is getting a lot of attention in the midst of the coronaviurs pandemic — the wearing of face masks and facial recognition technology. Mr. Simonite begins, “facial recognition algorithms from Los Angeles start-up TrueFace are so good, the U.S. Air Force to speed security checks through entraces to their bases/facilities. But, CEO Shaun Moore says he is facing a new question: How good is TrueFace’s technology when people are wearing face masks?”
“It’s something we don’t know yet, because it’s not been deployed in that environment,” Moore told WIRED. Not surprisingly, Mr. Moore told the publication that his company is “testing their technology on masked faces, and are hurriedly gathering images of masked faces to tune their machine-learning algorithms for pandemic times,” Mr. Simonite wrote.
“Facial recognition say [acknowledge] that facial recognition technology is generally less accurate when a face is obscured, whether by an obstacle, a camera angle or a mask – because there’s less information available to make a comparison,” Mr. Simonite wrote. But some vendors, as you might expect, have found creative ways to cope with identifying someone who’s face is obscured in some manner. “We can identify a person wearing a balaclava, a medical mask, or a hat covering a forehead,” said Artem Kuharenko, founder of NtechLab, a Russian company whose technology is deployed on 150,000 cameras in Moscow,” Mr. Simontie noted.
“But, Anil Jain, a professor at Michigan State University, who works on facial recognition and biometrics, said such claims are unverified,” Mr. Simonite wrote. Dr Jain added, “Companies can quote internal numbers, but we don’t have a trusted database yet,” to verify those claims.
Yuan Yang, writing in the March 18, 2020 edition of the Financial Times, noted that China-based Hanwang, a facial recognition compamy, boasts it has reached 95 percent accuracy in lab tests, in identifying face mask wearers. Huang Lei, the company’s Chief Technology Officer told the FT his system works for masked faces, by trying to guess what all the existing faces in the data base of photographs would look like if they were masked.
As Khari Johnson noted in the April 8, 2020 edition of VB, “COVID-19 is expected to change the world in significant ways, from telehealth, and video teleconferencing”; but, it may also lead to advancements in facial recognition technology that can identify someone — even if they are wearing a mask. And, last month, Northeastern University Professor Woodrow Hartzog and Evan Selinger from the Rochester Institute of Technology posted an April 6, 2020 article in the New York Daily News, “Masks And Our Face Recognition Future: How The Coronavirus (Slightly) Clouds The Future Painted By Tech Firms,” argued that “face masks are a temporary speed bump for facial recognition,” technology. “While they present a challenge,” Professor Hartzog does not believe that face masks in the age of COVID-19 will ultimately defeat inevitable advancements in surveillance technology.
Alas, advances in facial recognition, biometrics, artificial intelligence (AI) [data mining and machine learning are all advancing at the pace of Moore’s Law, or faster; and, it is clear that surveillance states like China will continue to advance the ball with respect to facial recognition and other surveillance means. The next generation of artificial-intelligence-driven facial recognition systems is already well underway.
Intrusive, oppressive surveillance is the autocrat’s new tool — see Richard Fontaine and Kara Frederick’s engrossing March 15, 2019 article in the Wall Street Journal, “The Autocrat’s New Tool.”
I suspect all is not lost; and this will be a constant cat-and-mouse game, with those who oppose this kind of intrusive surveillance coming up with new, clever and creative ways to hide their true identity. But, it is getting harder and harder to do so — as evidenced by what Communist China is already employing. How all this ends, is anyone’s guess. RCP, fortunascorner.com