Expert opinions, TECHNOLOGY

Masks, coronavirus and face recognition

Since the beginning of the century, we have witnessed at least three cross-border epidemics that influenced the public perception of all diseases: the 2002-2004 atypical pneumonia (SARS) outbreak, the 2003-2015 outbreaks of bird flu and swine flu, and the Middle East respiratory syndrome coronavirus (MERS), which appeared in 2012-2013 in Saudi Arabia and then spilled over to South Korea in 2015.

The Ebola virus epidemic in Africa stands apart; it was contained not least because the virus cannot be transmitted through the air.

One way or another, technological progress allows infectious diseases to spread rapidly around the world due to car, bus and train transit inside countries and air travel between countries and continents. The development of information technology makes it possible for the public to quickly learn about outbreaks.

We are currently witnessing the fourth epidemic, which surpasses the previous ones in terms of the number of confirmed cases, the speed at which it spreads, and the level of information support. The Chinese website that publishes official data about the spread of the 2019-nCOV coronavirus has about 4 mio views per hour, and the total number of views since the launch reaches 2 bln.

What do all these epidemics have in common? Face masks. Amidst the threat of a spreading virus, wearing medical masks and even respirators becomes a commonplace rather than something extraordinary. For example, this is a photo from a ballet class in Hong Kong from 2003:

It should be noted that wearing masks is becoming an absolutely normal phenomenon during an epidemic of a viral respiratory infection. Moreover, when there is a risk of catching the disease literally anywhere – on public transit, in a theater, in a store or at a gala event – wearing masks is making a shift towards being a trend and natural behavior. It becomes dangerous and almost unspeakable to leave home without a mask during an epidemic. By wearing a mask in public, we showcase our self-care by protecting ourselves from the risk of contagion, and care about others by not letting an incubating virus (anything is possible, right?) spread from us to other people. By doing so, we are not pursuing a goal of hiding our faces. But of course it happens.

But if everybody wears mask what about face recognition algorithms which, in the age of the Internet of Things are becoming if not essential but very important components of public security, targeted marketing and social interaction? Face recognition systems are being implemented on city streets, in airports, supermarkets, hotels, sporting events and other places that we are not even aware of. They are not only helpful in preventing unlawful activity by monitoring suspicious groups and individuals but are also contributing to targeting people with personalized ads.

Most face recognition algorithms require an image of uncovered face to identify a person. Such algorithms recognize facial features in their entirety; this is not about such methods as, say, iris scan for building access control. Upon spotting a medical mask or a respirator on a face, the recognition system will most likely refuse to identify a person. In case a system sees a mask with some neutral or texture image, such as abstract pattern, a jaw, or a Darth Vader helmet visor, the system is even more likely to refuse to identify the face. It is understood that while building and testing face recognition technologies their developers could not foresee scenarios where everyone in the street is wearing a mask. And this scenario has become reality.

It is clearly obvious that this reality renders the current face recognition systems useless and even somewhat dangerous as cases of false facial recognition may sharply increase in number thus leading to failures.

Face recognition systems will definitely be improved and their algorithms upgraded, and it is quite likely that soon we may receive an adequate solution that will allow facial recognition by the part of the face uncovered by mask and will produce a proper result for making decisions.

But it appears so far that the media hype over the coronavirus outbreak can not only ruin trans-border online trade markets built also have a dramatic effect on the spheres that have already started to get used to introduced facial recognition technologies as well as on the development of this technology.

By Vladimir Arlazarov, General Director, Smart Engines; Laboratory Head, Informatics and Management Federal Research Center, Russian Academy of Sciences

Previous ArticleNext Article