Rank One stands out in NIST test of image quality assessment for face biometrics
A benchmark of facial image quality assessment algorithms for preventing bad input from causing problems with face biometrics systems has been published by The National Institute of Standards and Technology.
The latest Face Recognition Vendor Test (FRVT) evaluated algorithms for scoring the quality of images used in border crossing. ‘Face Image Quality Assessment’ is part 5 of the ongoing NIST testing regime, with the first report published in July, and a second API and concept document following in August.
NIST evaluated 33 accurate face verification algorithms, at a false non-match rate (FNMR) threshold of 1 percent. The FNMR was recomputed after the lowest-quality 1 percent and 5 percent of samples were discarded to evaluate the assessment algorithms’ efficiency.
A pair of algorithms from Rank One Computing stand out as being among the most efficient in both scenarios. With the lowest-quality 1 percent removed, an ROC algorithm delivered an FNMR of 0.0059.
“Across the vast range of facial recognition use-cases it is critical to prevent ‘garbage’ from undermining the effectiveness of the system,” writes Rank One Chief Scientist, President and Co-founder Brendan Klare in a company announcement. “Whether it is border screening applications, online selfie enrollment for banking and enterprise services, law enforcement use cases, or just about any other use of face recognition technology, there is a substantial benefit for preventing face images that are of insufficient quality from ever being processed.”
Work on a standard for facial image quality assessment is ongoing, and the topic was the focus of a three-day workshop hosted by the EAB last November.