A study has found that face masks are causing significant issues with facial recognition software. The Verge has reported that the use of face masks is having a significant unintended effect for smartphone users.
Many tech companies have turned their attention to how they can help fight Covid-19. Such as FitBit which is working on a tracker to predict symptoms before they are present. There are also other examples of how apps and smartphones can help tackle the virus.
However, as more and more people are wearing face masks a new challenge has faced tech companies. Some estimates suggest error rates for facial recognition algorithms have jumped by up to 50%.
Facial recognition algorithms fail to deal with face masks
A study by the US National Institute of Standards and Technology (NIST) has found significant jumps on the error rates of the most popular facial recognition software. The spike was somewhere between 5 and 50% for most different algorithms.
It seems that black masks cause more errors than blue masks. It is also worse if the mask covers the nose compared just to the mouth. The latter is harldy surprising given its obscures more of the face.
Mei Ngan, an author of the report and NIST computer scientist pointed out the need for software to evolve in the current climate. She said, “We have begun by focusing on how an algorithm developed before the pandemic might be affected by subjects wearing face masks.”
Seh then went out to set out the plans for improvment. “Later this summer, we plan to test the accuracy of algorithms that were intentionally developed with masked faces in mind.”
This study confirms a number of anecdotal stories of facial recognition software struggling with face masks. The study only took into account one-to-one matching. This software operates in border crossings and passport control scenarios.
This is different to the type of algorithms used in mass surveillance, called one-to-many matching. Academic studies into one-to-many matching has not yet taken place.
The suggestion is though that if one-to-one matching is struggling then one-to-many will be as well. This is because one-to-many matching is more errror prone normally. Therefore, face masks are likely to exacerbate error rates.
Governments concerned about facial recognition errors
We have already heard the many governments are concearned about the potential impacts of face masks on facial recognitions software. However, for those that are advocates of more privacy this probably comes as welcome news.
Many tech companies are working on new algorithims to adapt to this new world. A Russian firm NtechLab claims it can identify individuals even if they’re wearing a balaclava. This comes from internal data so it is probably best to take this claim with a pinch of salt.
Expect tech companies to continue to work on this technology and quickly adapt to the new world. Ngan says “with respect to accuracy with face masks, we expect the technology to continue to improve.”
It is likely that we will see better software emerging over the next few months which is a testament to the adaptability of the technology sector. It is always interesting to see the unitened or unexpected consequences of the Coronavirus pandemic. Hopefully, companies will be able to come up with solutions in the near future.