ROC Ranks #1 in NIST Age Estimation Test
Computer vision firm ROC secured the top ranking for age estimation accuracy in a new evaluation from NIST. The company's algorithm was ranked #1 globally on both the Child Online Safety and Mugshot datasets, setting a new benchmark for biometric precision in digital safety.
The National Institute of Standards and Technology (NIST) Face Analysis Technology Evaluation (FATE) is a program that assesses facial analysis software, including age estimation. This evaluation is separate from the Face Recognition Technology Evaluation (FRTE), which focuses on identity verification. The FATE Age Estimation and Verification (AEV) track is an ongoing process where developers worldwide can submit their algorithms for performance evaluation on various datasets. The evaluation's primary metric is Mean Absolute Error (MAE), which measures the average absolute difference between the estimated age and the actual age. ROC's algorithm achieved the top global ranking in MAE for both the Child Online Safety (Ages 13-16) and the Mugshot datasets. These tests are crucial as facial age verification becomes increasingly mandated in legislation to protect minors online. The Child Online Safety dataset is particularly relevant as companies face more regulations to ensure online safety for children, requiring accurate age distinction around key age thresholds like 13, 16, and 18. The Mugshot dataset, officially known as NIST Special Database 18, contains 3,248 grayscale images and is used for testing identification systems in controlled environments. NIST's evaluations in this space have evolved, with the first report on automated age estimation performance published in 2014. The current FATE AEV track represents a more rigorous and ongoing benchmark compared to a decade ago, using a vastly expanded database of millions of photos from diverse sources to test algorithms. The demand for accurate age verification is driven by new laws in numerous U.S. states aimed at restricting minors' access to certain online content and services. This has led to social media platforms and financial institutions deploying more robust age assurance controls during onboarding and authentication processes. ROC, a U.S.-based company, develops its algorithms using what it describes as large, ethically sourced datasets to minimize bias and enhance precision. The company's technology is part of a broader platform that includes multimodal biometrics like face, fingerprint, and iris recognition, as well as real-time video analytics. Other companies actively participating and showing strong performance in NIST's age estimation evaluations include Innovatrics, Yoti, and Regula. The ongoing nature of the FATE AEV track allows for continuous benchmarking and reflects incremental improvements in the accuracy of age estimation technology from various global developers.
