SAN FRANCISCO, Dec 5 — An American study has shown that current facial recognition algorithms are powerful enough to identify people even if they are wearing a mask.
The results of today’s technologies are as good for people wearing masks as the best algorithms of 2017 were with mask-free faces.
This study was conducted by the US Commerce Department’s National Institute of Standards and Technology (NIST) over several months starting in spring 2020.
While the first results coming in during the summer were rather poor, with a 99 per cent error rate in certain cases, technology has considerably evolved to deliver reliable results.
The average error rate is currently around 0.3 per cent without a mask and around 5 per cent with one. In certain cases, masks can cover two-thirds of the face.
For their study, the researchers submitted over 6 million photos to 65 facial recognition algorithms.
They concluded that the latest performance rate was comparable to where technology was in 2017 on photos of people without masks.
The results of this study are all the more positive as masks have become increasingly widespread since the onset of the covid-19 pandemic, and are set to stay that way for several more months (at least).
The good news is that facial recognition tools won’t suffer from the use of masks in the public sphere and will continue to be able to give reliable results. — AFP-Relaxnews