Face Recognition Vendor Test Part 3: Demographic Effects
By Patrick Grother, Mei Ngan and Kayee Hanaoka, National Institute of Standards and Technology
* Facial recognition is a technology that has undergone a great deal of scrutiny --and received a good share of criticism --over the past decade or so.
Early versions were easily defeated and early adopters--such as police departments who wanted to identify most-wanted criminals as they strolled down the streets with cameras equipped with algorithms--had little success and ran into privacy concerns.
That hasn't stopped vendors from selling their wares. The National Institute of Standards and Technology recently evaluated 189 facial recognition products produced by 99 different firms to test their accuracy. Part one of the report found that overall, accuracy had been improving over the last five years. Part three looked at demographic biases.
So have these systems improved? The answer is "yes," if you happen to be a white male, according to this new NIST report. In other categories, females for example, or anyone who isn't white, the technology has shortcomings, it said.