ROC set new highs in NIST biometrics testing for age estimation

ROC believes it is well-positioned for success in the burgeoning market for biometric facial age estimation and verification, based on testing by the U.S. National Institute of Standards and Technology (NIST).

The age estimation and verification platform from ROC achieved the lowest Mean Absolute Error (MAE) among all participating vendors in the NIST Face Analysis Technology Evaluation (FATE), the company says in an announcement. ROC particularly improved in the 18-24 age segment, showcasing an MAE of 2.3 years using mugshot data.

ROC’s algorithms were validated against a vast dataset of 11.5 million photos in the NIST evaluation. This dataset included mugshots, border-crossing webcam images, and immigration headshots of individuals from over 100 countries.

The company says that the comprehensive training of its models demonstrates its capacity to accurately estimate ages across diverse demographics, including various human lifespans, geographic origins, and genders.

Its AI-powered automated age verification system offers businesses a way to enhance accuracy by allowing only verified customers of legal age to access age-restricted services. Implementing this system can increase cost efficiency by minimizing the need for manual oversight, which is often time-consuming and susceptible to human error.

The demand for age estimation and verification is expected to rise due to new legislative changes in the United States. Currently, 18 states, including Alabama, Arkansas, and Florida have implemented age verification laws to protect minors from accessing inappropriate content and services.

How ROC develops its deep learning algorithms

ROC also provided information on the development of their deep learning algorithms. Their software utilizes computer vision and biometric analytic algorithms across various modalities, such as face recognition, fingerprint matching, object detection, and license plate recognition.

The development of deep learning algorithms involves four key steps: data development, algorithm development, algorithm integration, and customer support.

The process begins with data development, which involves building and validating the training data. This training data is then utilized to develop algorithms. The trained algorithms are tested and evaluated for both absolute accuracy in real-world scenarios and relative accuracy by comparing the performance of different algorithms.

After developing the algorithms, the next step involves software integration, where the model is ported into deployable software libraries. Once the integration is completed, extensive testing is conducted to validate the integration and ensure the computer vision models perform as expected in their integrated form. Finally, direct communication is established with integrators and customers to gather feedback on the algorithms’ performance in the field.

In addition to its success with face biometrics, ROC has recently scored strong marks in an evaluation of latent fingerprint biometrics matching by NIST this year.

Article Topics

age verification  |  biometric age estimation  |  biometric testing  |  biometrics  |  computer vision  |  deep learning  |  face biometrics  |  facial age estimation (FAE)  |  ROC