Facial recognition software designers stand to win $25,000 in the Intelligence Advanced Research Projects Activity’s latest challenge.
IARPA, in partnership with the National Institute of Standards and Technology, recently launched its Face Recognition Prize Challenge. The agency is seeking software to improve core facial recognition accuracy. According to the challenge website, the goal is to be able to identify individuals in the “wild,” meaning those who have been photographed in various conditions and environments.
Many people have formal photos on record through their passports and driver’s licenses. However, IARPA’s challenge is to explore technologies that can analyze “noncooperative” images in which the subject is unaware that the picture is being taken. Noncooperative recognition requires unknown video imagery to be compared with images previously collected from multiple individuals.
A demanding blend of time, cost, and manpower are needed to process thousands of facial images, according to NIST’s Face In Video Evaluation (FIVE) program report, which was released March 7. The FIVE program, established in 2014, tests how well different facial recognition algorithms identify people appearing in video sequences.
“The report argues that this is complicated because there a number of different factors,” said Patrick Grother, one of the NIST scientists behind the report, in an exclusive interview with MeriTalk in April. “If it’s to be a success, you need a multidisciplinary team. You might change the environment to try to arrange for subjects to look at the screen. It’s not one-size-fits-all.”
People who are interested in the IARPA challenge are instructed to send precompiled software libraries to NIST, where scientists will test the algorithms against reams of stored images. Academics, industry representatives, Federal employees, and members of research organizations are invited to apply.
The producer of the most accurate one-to-many identification software will win $25,000. The producer of the most accurate one-to-one verification software will win $20,000; $5,000 will go to the group that creates the fastest one-to-many identification software.
The facial recognition challenge is not the only identification technology contest IARPA is hosting. Chris Boehnen, an IARPA program manager who will be judging the facial recognition submissions, is also leading the Nail to Nail Fingerprint challenge. The Nail to Nail challenge focuses on fingerprinting technology.
Participants have until June 15 to submit their algorithms for the face recognition challenge.