Facial Recognition Tech Can Recognize Masked Faces

This post is also available in: עברית (Hebrew)

A new deep learning facial recognition technology has been unveiled recently. Jointly developed by Panasonic Business and the National University of Singapore, Face Server’s deep learning core engine is the industry’s most accurate yet, tests by the National Institute of Standards and Technology (NIST) have found.

The technology can identify faces at an angle of up to 45 degrees to the left or right or 30 degrees up or down, as well as those partially hidden by sunglasses and face masks.

The platform has already been rolled out at Tokyo’s Haneda airport, replacing a fingerprint recognition system at the passport entry gates after three years of verification testing.

Among the platform features: Real-time processing capacity of up to 20 cameras per server and can execute high-speed searches of up to 30,000 registered reference faces.

iA function means image analysis is performed on the camera and only the best images are sent to the server – reducing server and network load and total system cost, according to ifsecglobal.com.

“Face Server changes the game in terms of detection accuracy,” said Gerard Figols, Category Manager at Panasonic Security. “Not only does it maintain 90%+ accuracy for faces that are partially covered by sunglasses or face masks, it also recognizes faces from photographs that are up to 10 years old.”