Before the pandemic, the fingerprint scanner was gaining popularity as a new way to identify ourselves to authorise payments or clock into offices. Now, we can expect that the technology will be dropped in favour of a contactless one – facial recognition.
But facial recognition technology relies on seeing our whole face, or as much of it as possible. Those who have used the tech know how sensitive it can be to small changes such as how we wear our hair or whether we are wearing glasses or not.
Now, much of the world is mandated or otherwise persuaded to wear a mask in public. This undoubtedly poses a big problem – can facial recognition technologies still identify us when half of our face is covered?
Apparently, some technologies have already adapted. In Asia, one company in Japan is one of many claiming to be able to identify faces wearing masks. Apple also claims its Face ID will be able to recognise faces wearing masks with its upcoming iOS 13.5 update.
But are these technologies providing the same experience as they can with uncovered faces? Facial recognition expert Eric Hess points out in Security Magazine that as facial recognition technology has been rolled out, their databases have grown, and with this more faces that look like us have been added.
What sets us apart from others who look like us could be features concealed by a mask – our noses, mouths, jawlines. The quality of the images in the database, including the lighting, can further blur the distinction between us and people who look like us.
“When you have fewer than 100,000 people in the database, you will not feel the difference,” Alexander Khanin, CEO and Co-Founder of VisionLabs, a machine learning startup told Wired early this month. However, with one million people, Khanin says accuracy will be noticeably reduced.
But companies are failing to admit the drawbacks. “Companies can quote internal numbers, but we don’t have a trusted database or evaluation to check that yet,” says Anil Jain, a professor at Michigan State University. Facial recognition companies must start to acknowledge these limitations. If they do not, misconceptions could lead to inappropriate rollout of the technology.