Tattoos are unique—even when the concept is unoriginal, in practice a tattoo still ends up with it own foibles and traits. According to Nature, the US government organized a competition to be able to use those artsy designs to better identify criminal suspects.
Its security agencies want to be able to identify suspects from images using computer vision. For instance: If security cameras have captured a person robbing multiple banks, the government would like an algorithm that could determine that a tattoo that appeared on each bank-robbing suspect was the same tattoo. It could also help determine if certain tattoos are tied to gang members, Nature reported.
“They need to know, does this person represent a threat, and they need to know that really quick,” says Daniel Olson, an FBI analyst, told Nature.
The competition, which was held on June 8, was partially sponsored by the FBI, the Department of Justice and the Department of Defense. The National Institute of Standards and Technology (NIST), which organized the competition, told Quartz that the meeting was “a preliminary research challenge much different, from our more stringent competitions.” There was no prize at stake, and the challenge “provided input for discussion of what approaches the biometric community (researchers and agencies) should pursue.”
The six teams that entered the competition—from universities, government entities, and consulting firms—had to develop an algorithm that would be able to detect whether an image had a tattoo in it, compare similarities in multiple tattoos, and compare sketches with photographs of tattoos. Some of the things the National Institute of Standards and Technology (NIST), the competition’s organizers, were looking to interpret in images of tattoos include swastikas, snakes, drags, guns, unicorns, knights, and witches. (Prospective gang members seem to have a thing for fantasy fiction.)
Computer vision, the ability for computers to interpret objects in images, is still in its infancy. As it stands our most advanced computers struggle to tell images of cats apart. Many companies—including Google—and universities are trying to solve this problem, and the implications extend far beyond detecting persons of interest in police investigations. As Stanford University’s director of artificial intelligence research, Fei-Fei Li, suggested in her TED talk, in the future, we could have security systems that can actually respond to footage of a child drowning by detecting what is happening, rather than just recording it.
NIST said that it hopes to publish the findings from the challenge within the next six months, but it doesn’t expect to have figured out computer vision any time soon, meaning this competition will likely have no real winner.