NIST biometric age estimation update show demographic, accuracy gains

Demographic disparities and mean error rates are falling among the newest age estimation and verification algorithms submitted to the U.S.’ NIST for analysis,

The first update since February to the Face Analysis Technology Evaluation (FATE) Age Estimation & Verification includes new algorithms from Dermalog (003, meaning the fourth entry from the company), Innovatrics (003), Cognitec (001) and a debut of the Onfido (000) algothim submitted by Entrust.

As of the May 15 report, Dermalog holds the lowest false positive rate (FPR) in the Challenge 25 application, at 0.017 +/- 0.010.

Innovatrics scored top-five results for age verification in the latest update, and even stronger in Challenge 25 scenarios. The company also scored low mean absolute error (MAE) rates for East Asian, Eastern European and South East Asian males while reducing its MAE for East African males and females below the 3.5 threshold at which NIST colors results yellow.

Cognitec cracked the top 5 in Challenge 25 scenarios, MAE for Child Online Safety (13-16), while cutting its MAE for East African females.

Entrust’s Onfido algorithm delivered a false negative rate (FNR) of 0.060 for females and 0.128 for males, lowest and second-lowest yet in FATE Age, respectively.

Algorithms from regular leading performers Yoti, ROC, Regula and Idemia remain among the best-performing across various metrics.

Article Topics

age verification  |  biometric age estimation  |  biometric testing  |  biometrics  |  Cognitec  |  demographic fairness  |  DERMALOG  |  Entrust  |  Face Analysis Technology Evaluation (FATE)  |  Innovatrics  |  NIST