BANDAR SUNWAY – Facial recognition is a way of identifying or confirming an individual’s identity using their face, and it is a category of biometric security. Facial recognition systems can be used to identify people in images, videos, or in real-time. This identification typically functions like a face scanner and is used to access an application, system, or service. Currently, there is growing interest in using the technology in other fields, but security and law enforcement still make up for the majority of its uses.
In our previous article, we touched on what the NIST Face Recognition Vendor Test (FRVT) is and what it is used for. To recap, the FRVT provides independent evaluations of commercially available and prototype face recognition technologies. Achieving top rankings on the FRVT proves that a developer’s facial recognition algorithm provides accurate and secure results.
The FRVT project is now ongoing, with regular reports, and its scope is expanding. As of now, it has included tests for facial morphing detection and testing for demographic effects (e.g., age, gender, and race). Below, we have listed down all the current FRVT activities for which developers can still submit their algorithms.
FRVT: Face Mask Effects
On November 30, 2020, NIST published NISTIR 8331, which is aimed at quantifying face recognition accuracy for people wearing masks. This report adds:
Their initial approach has been to apply masks to faces digitally, which allowed them to leverage the large datasets that they already have. This report quantifies the effect of masks on both false-negative and false-positive match rates.
FRVT: Demographic Effects
In the NISTIR 8280 report, NIST describes and quantifies demographic differentials for contemporary face recognition algorithms. They have tested over 200 face recognition algorithms from nearly 100 developers to determine the demographic disparities. Four photo collections totalling more than 18 million images of more than 8 million individuals were used in the tests.
FRVT 1:1 Verification
The FRVT Ongoing activity is conducted on a continuing basis, so this allows developers to submit their algorithms to NIST whenever they are ready. To assess the effectiveness of face recognition algorithms created in commercial and academic institutions around the world, very large sets of facial imagery will be used in the evaluation. Many face recognition-related evaluation tracks will be run under this test.
FRVT 1:N 2018
The FRVT 1:N 2018 will assess improvements in one-to-many face recognition identification algorithms’ accuracy and speed when exploring enrolled galleries with at least 10 million IDs. The evaluation will primarily use standardised portrait images and will quantify how accuracy depends on subject-specific demographics and image-specific quality factors.
Facial morphing and the ability to detect it is an area of high interest to photo-credential issuance agencies and those employing face recognition for identity verification. Prototype facial morph detection systems will be continuously and independently tested as part of the FRVT MORPH test.
NIST is evaluating algorithms for assessing the quality of facial images. Large collections of photos will be subjected to quality assessment algorithms, and their results will be compared to the results of face recognition.
Previous Evaluations in the FRVT Series
Among the previous evaluations that have been done in the FRVT series were in 2000, 2002, 2006, 2010, 2013, and 2017.
About WISE AI
WISE AI is an award-winning Artificial Intelligence company specialising in digital identity technologies. We develop world-class emerging deep tech that is adopted by the government and multiple industries. Our AI-powered solutions include EKYC, digital ID, digital signature, and blockchain. Our technology is optimised for the recognition of ASEAN faces.