The National Institute of Standards and Technology (NIST) is transforming its renowned Face Recognition Vendor Test (FRVT) into two distinct evaluation programs. One of them, the “Face Recognition Technology Evaluation” (or “FRTE”) will be dedicated to algorithms that “involve the processing and analysis of image”, whereas the other, “Face Analysis Technology Evaluation” (“FATE”) will concern identity verification specifically.
Essentially, the aim is to distinguish between facial recognition technologies and face biometrics technologies that aren’t used for biometric matching, such as age estimation tools.
The FRVT program was born in the early 2000s when NIST recognized the growing importance of facial recognition technology. At that time, facial recognition was emerging as a powerful tool with applications ranging from security and law enforcement to digital identity verification and access control.
In its nascent stages, the FRVT program conducted foundational evaluations of facial recognition algorithms. These assessments focused on measuring the accuracy and reliability of these algorithms in performing verification (one-to-one matching) and identification (one-to-many matching) tasks. These evaluations not only helped benchmark the state of the art but also drove improvements in the technology.
As facial recognition technology advanced, so did the FRVT program. NIST introduced regular updates, new benchmarks, and datasets to ensure that evaluations reflected real-world challenges. This iterative approach enabled the program to keep pace with the rapid evolution of facial recognition algorithms.
In response to the increasing relevance of facial recognition in diverse applications, NIST broadened the scope of the FRVT program. Evaluations began to cover various aspects of facial recognition, including the processing of still images and video-based facial recognition systems. This expansion allowed the program to address the unique challenges posed by different scenarios.
In recent years, NIST recognized the importance of fairness and bias assessment in facial recognition technology. The program undertook assessments to measure how different algorithms performed across demographic groups.
Now, by splitting the FRVT program into two distinct tracks, NIST is poised to help add clarity to its independent evaluations. In so doing, it could help to give those outside of the biometrics industry a better understanding of the differences between facial recognition technology and analogous technologies like age estimation and emotion detection, which tend to play very different roles in real-world deployments.
September 11, 2023 – by the FindBiometrics Editorial Team