Facial recognition algorithm disrupted by face masks; says a recent government study

Original article was published by Quatics on Artificial Intelligence on Medium


Facial recognition algorithm disrupted by face masks; says a recent government study

Face masks being one of the top defenses against the spread of coronavirus has become a part of every household. Be it going to the grocery store, work, schools, and colleges, or even to the apartment gate, face masks have become a necessity during this pandemic.

Though face masks have become extremely beneficial, their growing use is having a second, unintended effect: disrupting facial recognition algorithms.

Photo by Macau Photo Agency on Unsplash

According to a study by the US National Institute of Standards and Technology (NIST), since face masks cover adequate areas of the mouth and nose, the error rate on facial recognition algorithms escalates from 5 percent to 50 percent.

The study stated that the more mask coverage, the higher the error rate since the algorithms found it difficult to identify the face, and that black masks gave higher error rates than blue masks.

Photo by Gayatri Malhotra on Unsplash

We have started by focusing on how an algorithm developed before the pandemic might be affected by subjects wearing face masks. We have developed facial algorithms intentionally with masked faces in mind, and Later this year, we have planned to test the accuracy of these algorithms” said Mei Ngan, an author of the report and National Institute of Standards and Technology computer scientist.

The algorithms tested by NIST work by measuring the distances between features in a user’s face, and since these masks reduce the accuracy of these algorithms by eliminating most of these features, the errors are prominent.

The type of facial recognition that was tested in the NIST’s report is known as one-to-one matching. This is used in border crossings and passport control scenarios to ensure whether a person’s face matches their ID. The other type is known as a one-to-many system, which is used for mass surveillance, where a crowd is scanned to find matches with faces in a database.

Though NIST’s report only included one-to-one matching, the other type is more prone to such errors. Therefore, if face masks are breaking one-to-one systems, they’re likely breaking one-to-many algorithms with probably a greater frequency.

Due to the current scenario, companies building facial recognition tech have started rapidly adapting to this new world, designing algorithms that identify faces by just using the area around the eyes.

“We are planning to test specially tuned facial recognition algorithms for mask wearers later this year and along with that we aim to probe the efficacy of one-to-many systems,” says NIST.