Kneron compares facial biometric algorithm favorably to lightweight AI models in NIST testing
Kneron has scored the best aggregate rating among all lightweight biometric facial recognition models under 64MB in the NIST Face Recognition Vendor Test (FRVT) 2019.
There were six models submitted for the NIST FRVT 2019 – Verification under 64MB, and the Kneron-003 algorithm, which takes 57MB, outranked them all in aggregate scoring. Among models taking up less than 100MB, Kneron-003 placed fifth in the VISA category and second in the mugshot category, according to the company.
The test results show that its pioneering artificial intelligence models can enable mass adoption of on-device edge AI, the company says. With a focus on developing technology and lightweight algorithm models balancing fast computing speeds and low power consumption, Kneron’s facial recognition can be applied in small and mobile form factors, such as security cameras, drones, and smart locks.
The company also says its algorithm scored higher in some sections of the test than submissions from renowned AI companies with models larger than 1GB.
Kneron partnered with Etron Technology at the beginning of the year to develop 3D facial recognition at the edge using natural light.