Mask and you shall receive: New facial recognition technology can recognise faces even whilst wearing a mask

  • Anton Nazarkin, , Global Business Development Director at VisionLabs

  • 04.03.2022 06:45 pm
  • #contactless

The pandemic has altered how we live, making face masks commonplace and the desire for contactless technology in all forms more prevalent. We have all gone through access control gates on the subway or at live events, and experienced the inconvenience of pulling down our mask to be scanned. This is because many companies have not yet deployed technology smart enough to scan a face with a mask. For consumers who demand a seamless, contactless experience, this process has become extremely frustrating.

The problem is that many facial recognition softwares fail liveness checks when a mask is present, due to the inability to see a whole face - or worse they will validate a fraudulent scan due to a low number of data points being scanned and poor liveness checks.

Most vendors can't have high similarity scores (a similarity score is a statistical measure of how likely two faces in an image are the same person) in face recognition when the mask is on, simply because the face is less visible to provide strong confidence levels in face recognition. However VisionLabs, have managed to overcome this challenge by having the highest levels of accuracy and world leading liveness checks.

Behind the scenes

One key element of accurate facial recognition is the liveness check. The liveness check ensures a real, living human is in front of the camera and not a printed photo or a video of a face. There are several types of liveness checks. The cooperative/active liveness check is just one of these, however it is an outdated technology which can be easily spoofed. VisionLabs utilizes one shot liveness, which checks for liveness without the object needing to interact (by blinking for example) with the liveness checking system. The one shot liveness is more user friendly, as there is no need for additional actions, whilst also enhancing the reliability of mobile and web authentication and reduces the possibility of fraud.

VisonLabs facial recognition technology also incorporates a depth liveness check. This allows the technology to assess how far images being scanned are - down to the millimeter - essentially meaning the technology “understands” if the image is flat (like a photo or video of a face) or whether it is an actual face. VisionLabs one shot liveness is unique and allows us to use Red Green Blue Color Model (RGB) for liveness check reliably. The near-infrared (NIR) is another layer of protection, making it much more reliable than simple cooperative/active liveness.

Therefore, even when an individual is wearing a mask, the VisionLabs technology can recognize these faces to a much higher standard compared to most facial recognition softwares.

VisionLabs software was rated the highest in the NIST report*, with an accuracy of 99.9% for both 1:1 and 1:To Many, meaning that when a mask is present the level of accuracy is still superior and best in class.

In real time

What this means for the customer, the commuter or the employee is that they are able to face the device with the comfort of safety as they still wear a mask. Within milliseconds their face is scanned and they are provided with a ‘successful’ screen. New habits around masks have been formed globally and while it is unlikely that wearing a mask outdoors will become the norm, it is likely that it will remain common to wear a mask in cramped and crowded places such as concert venues, subways and on airplanes.

In this future that none of us could have predicted, businesses need to use the right technology to make our experiences as seamless as they once were. It will benefit the customers and employees, and themselves. The benefits of this technology to businesses are numerous: a more secure operation and less fraud means more efficient processes and less costs, whilst greater accuracy means better customer/employee experience.

*VisionLabs’ facial recognition algorithms have been more consistently ranked in the top few by the U.S. National Institute of Standards and Technology for its accuracy than any of its competitors, as tested by the National Institute of Standards and Technology.

Other Blogs