Mandatory Face Masks Are Thwarting Facial Recognition In China
Competing public policy imperatives can be so inconvenient. JL
Anne Quito reports in Quartz:
As millions don masks across the country, the Chinese are discovering that face
masks trip up facial recognition-based functions, a technology necessary
for many routine transactions in China. Mobile
phones, condominium doors, and bank accounts won’t unlock with a glance. Most complaints are about unlocking mobile devices. Apple confirmed anunobstructed view of a user’s eyes, nose, and mouthis needed for FaceID to work. The technology is a linchpin in the Chinese government’s scheme topolice its 1.4 billion citizens. Anyone registering a new mobile phone SIM card(must) undergo a face scan.
Face masks are mandatory inat least two provincesin China, including the city of Wuhan. In an effort to contain the coronavirus strain that has causednearly 500 deaths, the government is insisting that millions of residents wear protective face coveringwhen they go out in public.
As millions don masks across the country, the Chinese are discovering an unexpected consequence to covering their faces. It turns out that face masks trip up facial recognition-based functions, a technology necessary for many routine transactions in China. Suddenly, certain mobile phones, condominium doors, and bank accounts won’t unlock with a glance.
Complaints are plentiful in the popular Chinese blogging platform Weibo, reports Abacus, the Hong Kong-based technology news outlet. “[I’ve] been wearing a mask everyday recently and I just want to throw away this phone with face unlock,” laments one user. “Fingerprint payment is still better,” writes another. “All I want is to pay and quickly run.”
Most complaints are about unlocking mobile devices. Apple confirmed to Quartz that anunobstructed view of a user’s eyes, nose, and mouthis needed for FaceID to work properly. Similarly, Huawei says that its efforts to develop a feature that recognizes partially-covered faces has fallen short. “There are too few feature points for the eyes and the head so it’s impossible to ensure security,” explains Huawei vice president Bruce Lee, ina Jan 21 post on Weibo.”We gave up on facial unlock for mask or scarf wearing [users].”
Biometrics, including facial recognition, are essential to daily life in China, on a scale beyond other nations. It’s used to do everything from ordering fast food meals to scheduling medical appointments to boarding a plane in more than 200 airports across the country. Facial recognition is evenused in restroomsto prevent an occupant from taking too much toilet paper.
Hong Kong’s government, incidentally, has been trying to install a“mask ban”for protestors participating in anti-government rallies. The anonymity afforded by surgical masks, gas masks, and respirators has somehow emboldened both police and demonstrators to act aggressively, without fear of being caught on camera
Facial recognition technology that can “see through” disguises already exists, but it’s far from perfect. Researchers at the University of Cambridge and India’s National Institute of Technology, for instance, demonstratedone methodthat could identify a person wearing a mask with around 55% accuracy. In 2018, Panasonic introducedcommercially-available softwarethat can ID people wearing surgical masks if the camera captures images at a certain angle.
Despite its widespread adoption across China, it’s ironic that facial recognition technology in general has been found to beless reliable when processing non-white faces, observes Jessica Helfand, author of the new bookFace: A Visual Odyssey. “The fact that surveillance is increasingly flawed with regard to facial recognition and Asian faces is a paradox made even more bizarre by the face mask thing,” Helfand says.
A recent landmark study by the US National Institute of Standards and Technology revealeda racial bias in algorithmssold by Intel, Microsoft, Toshiba, Tencent, and DiDi Chuxing. It showed that that African Americans, Native Americans, and Asians were 10 to 100 times more likely to be misidentified compared to a Caucasian subject.
As a Partner and Co-Founder of Predictiv and PredictivAsia, Jon specializes in management performance and organizational effectiveness for both domestic and international clients. He is an editor and author whose works include Invisible Advantage: How Intangilbles are Driving Business Performance. Learn more...
0 comments:
Post a Comment