Join or Sign in

Register for your free membership or if you are already a member,
sign in using your preferred method below.

To check your latest product inquiries, manage newsletter preference, update personal / company profile, or download member-exclusive reports, log in to your account now!
Login asmag.comMember Registration

NIST says masks hurt face recognition technology. SAFR responds

NIST says masks hurt face recognition technology. SAFR responds
A report from NIST says that face masks could cause as much as a 5-50% error rate in facial recognition technology. SAFR, a major vendor in this segment responds to this.
In July, the US National Institute of Standards and Technology (NIST) noted that its study found face masks posed significant challenges to facial recognition technology, with a 5-50 percent error rate. Mainstream media reported this finding, which contradicted claims from certain large manufacturers and created confusion among customers.

This is extremely relevant as more and more companies try to enter this market with AI facial recognition solutions, given the increased demand that COVID-19 has prompted. NIST FRVT (Facial Recognition Vendor Test) is considered an industry benchmark, and any report from the same agency has a significant impact on the market.

What the facial recognition tech companies say

One of the companies that claim accurate facial recognition despite masks is SAFR. recently spoke to Eric Hess, Senior Director of Product Management at SAFR to get their comments on the report.

“The NIST evaluation published in NISTIR 8311 is specifically an evaluation of the capabilities of Face Recognition algorithms developed BEFORE the COVID-19 pandemic, and their ability to handle faces wearing masks, which is a new social norm,” Hess said. “To quote the report, ‘This report is intended to support end-users to understand how a pre-pandemic algorithm might be affected by the arrival of a substantial number of subjects wearing face masks. The next report will document accuracy values for more recent algorithms, some developed with capabilities for recognition of masked faces.’

Hess used an analogy to explain this further, adding that to put it in a COVID-19 context, it would be like evaluating why Penicillin is not an effective treatment for COVID-19. Penicillin was around long before COVID-19 and developed for a different set of illnesses, well before the novel Coronavirus-19 strain.

“Based on the analogy presented, it is no surprise that technology pre-dating an emerging shift in societal behaviors should be evaluated as being lack-luster in its ability to address the new challenges,” Hess said.

Do masks negate the benefits of facial recognition technology?

The main question that the NIST report has raised is if facial recognition technology becomes redundant after COVID-19 because more and more people use face masks. Hess disagrees with the idea that face masks negate the benefits of facial recognition, pointing out that’s its too broad a statement.

“It would be like stating the current uncurable status of COVID-19 negates the benefit of Medicine,” Hess said. “The 2019 state of medicine is incapable of treating every COVID-19 patient; we can all agree on that. But doctors and researchers in the medical field are making headway in this area, benefiting from previous anti-retroviral advances in the era of AIDS and their current-day dedication to addressing this new medical issue.”

Medicine and Technology share similarities in this regard, Hess noted. They are practices that evolve over time. Through dedicated research, testing, trials, and diligent benchmarking and evaluations, we can develop new approaches to address societal challenges, whether they be of a viral challenge to medicine, or a behavioral challenge to computer vision.

“A more technical description of why masks do not negate the value of face recognition is because our ability to ‘tease-out’ uniquely differentiating characteristics in the periocular region of the face is rapidly advancing,” Hess said. “Depending upon the size of DB needed and the specific FR workflows supported, we can easily deliver high-quality face recognition with only the region around the eyes and eyebrows visible. The advances in this area are a mix of training, data availability on which to train, and the increased prevalence of sensors (camera technology) that have the ability to capture higher resolutions and faster shutter speeds, so image clarity, motion blur, and accurate representation of real-world data is possible, and advancing rapidly.”

The bottom line is that it may be correct to say that face masks negate the benefits of face recognition algorithms that pre-date the widespread usage of face masks. However, post COVID-19 outbreak, much research has gone into making facial recognition technology efficient on people wearing masks. It may not be perfect yet but cannot be considered ineffective either. That said, knowing how to select the best solution may make the all the difference to the customer. 

Product Adopted:
Subscribe to Newsletter
Stay updated with the latest trends and technologies in physical security

Share to: