Innovatrics is assuring customers that its facial recognition solutions are still highly accurate when identifying people who are wearing masks. The company issued the statement in response to the COVID-19 pandemic, which has prompted many people to wear face masks in an effort to reduce the threat of transmission and slow the spread of the virus.
“As soon as masks started becoming commonplace we began collecting test data to see how face recognition would be affected,” said Innovatrics Solution Manager Metod Rybar. “We found that even without any changes to the settings, the system is still 95 percent accurate when recognizing a given person in a dataset of hundreds of faces.”
While the system is 95 percent accurate as is, Innovatrics claims that customers can boost that number to 99 percent with a few small adjustments. Most notably, they can have two photos for each individual in their database, one with a mask and one without. Customers can set different thresholds for each photo using a feature that already comes with SmartFace and other Innovatrics facial recognition solutions.
Though the effect is minimal, the Innovatrics system is slightly less accurate when a mask is involved. As a result, organizations with extremely high security requirements may need to conduct more manual checks on people who are wearing masks.
The Innovatrics statement comes nearly a month after Herta launched a new version of its facial recognition algorithm that was designed specifically for a masked environment. The Chinese company Hanvon has also claimed that its solution is 95 percent accurate when dealing with masks. Innovatrics’ SmartFace, meanwhile, was recently integrated into the Milestone XProtect video management software platform.
April 6, 2020 – by Eric Weiss