Software of the American startup ClearView AI entered the top 10 of nearly 100 providers of face-recognition technology in a test by the U.S. National Institute of Standards and Technology (NIST).
According to the report, the ClearView AI system performed worse in another version of the test that simulated the use of biometric authentication to secure access to buildings, for example, identifying employees.
Last month, the company demonstrated satisfactory results in the so-called individual test — the ability to match two different photos of the same person, simulating facial verification such as FaceID to unlock a smartphone.
According to the CEO of the company, Hoan Ton-That, the positive results “opened new opportunities for the sales team”.
SenseTime, a Chinese startup, topped the test. The top ten also included solutions from Russian companies VisionLabs and NtechLab.
NIST has conducted face-recognition supplier tests for two decades. The report’s authors note that during this time the technology has undergone a revolution: algorithms have become more tolerant of poorly lit scenes and low-quality images, as well as to “poorly posed” objects.
Earlier, Ton-That said that the ClearView AI database grew to 10 billion images. The company is also developing a facial-recognition system for masks.
In November, Australian authorities demanded that ClearView AI cease collecting citizens’ photographs and destroy the data it already holds about them.
In April, a U.S. senator proposed to legally prohibit government agencies from buying biometric data from private companies without a court order.
In February, Canadian regulators accused Clearview AI of unlawful data collection about citizens and “recommended” halting services in the country.
Subscribe to ForkLog news on Telegram: ForkLog AI — all the news from the world of AI!