For the latest news and information on the coronavirus pandemic, visit the WHO website.
Many facial recognition companies have claimed that they can pinpoint people even with face masks, but the latest results from a study show the covers dramatically increase error rates.
In an update on Tuesday, the U.S. National Institute of Standards and Technology examined 41 facial recognition algorithms submitted after the COVID-19 pandemic was declared in mid-March. Many of these algorithms were developed with face masks in mind and claimed that they could still accurately identify people even when half of their face was covered.
Track the coronavirus pandemic.
In July, NIST released a report that found that face masks thwart regular face recognition algorithms with error rates between 5% and 50%. NIST is widely recognized as the leading authority on the testing of facial recognition accuracy and expects algorithms to improve the identification of people in face masks.
That day is yet to come, as each algorithm saw at least minor increases in error rates as masks came into the picture. While some algorithms were still overall accurate, like Chinese facial recognition company Dahua's algorithm error rate from 0.3% without masks to 6% with masks, others had error rates that rose to 99%.
Rank One, a facial recognition provider used in cities like Detroit, had an error rate of 0.6% without masks and an error rate of 34.5% when masks were applied digitally. In May, the company began "periocular detection," which claimed to be able to identify people right in front of their eyes and nose.
Rank 1 CEO Brendan Klare said the company was unable to submit this algorithm to NIST because the agency is limited to one submission per organization.
"Therefore, the NIST mask study does not reflect our ability to perform identification in the presence of masks," Klare said in an email.
With TrueFace, used in schools and at air force bases, the algorithm's error rate increased from 0.9% to 34.8% after adding masks. The company's CEO, Shaun Moore, told CNN on Aug. 12 that its researchers were working on a better algorithm for detection beyond masks.
The companies did not respond to a request for comment.
While each face detection algorithm had a higher error rate after adding masks, some error rates were as low as 3%, indicating that it is not impossible for algorithms to identify people even when their faces are covered.
Face masks are proven tools for limiting the spread of the novel coronavirus. Governments around the world have mandated that people wear covers to lessen the effects of the outbreak. Health professionals expect the majority of people to have to do this Keep wearing masks for yearsPush facial recognition companies to improve their algorithms.
NIST keeps reporting on how masks have affected facial recognition algorithms. 6 million images from his database were used and a mask was digitally added to the photos.
It is possible that error rates are higher when NIST uses real photos of people in masks instead of digitally added cover, as physical masks can have different shades, textures, and patterns that also confuse algorithms.
The information contained in this article is for educational and informational purposes only and is not intended as health or medical advice. Always consult a doctor or other qualified health care provider with any questions about a disease or health goals.
