It wasn’t just Reddit that failed last week in its quest to find the Boston bombing suspects — facial recognition technology also reportedly failed to produce a match for the suspects.
This certainly stands in contrast to hype about the technology last week.
“Facial recognition cuts investigation time down to seconds,” ran one headline on CNET, while Businessweek reported that the technology likely played a role, and law enforcement officials are “relying on [it] more and more”.
Proponents of the technology say the apparent failure isn’t with facial recognition technology itself, however, that instead it was the limitation of CCTV technology that prevented facial recognition from matching the suspects.
Most of the information gathered against the subjects appears to have come from more traditional sources.
As Asjylyn Loder & Esmé Deprez wrote last week for Bloomberg, one key factor in initially identifying the two suspects was an old-fashioned witness. Jeff Bauman, a spectator who had both his legs amputated, alleged that the older suspect (Tamerlan Tsarnaev) place a bag down momentarily before the explosions.
“Bag, saw the guy, looked right at me,” he wrote when he first woke up, later describing Tamerlan so that investigators knew what to look for in the hours of video they had.
After looking over these videos, investigators apparently found something that clearly linked the two suspects to the bombing — Massachusetts Gov. Deval Patrick has said that the FBI has “chilling” video not released publicly that places Tamerlin right at the scene of the attack.
Despite CCTV images of two suspects, facial recognition software still did not play a role, Boston Police Commissioner Edward Davis told the Washington Post — even though both suspects’ images were in official databases facial recognition software did not provide their names.
Instead, the next step was human — several people came forward to suggest that the Tsarnaev brothers looked like the suspects (including, reportedly, their aunt). These identities were confirmed by a driving licence found on the body of Tamerlan after he engaged in a shoot out with police officers.
One problem for facial recognition technology was likely the image quality of the CCTV images. According to CNBC, the standard for accurate facial recognition is 90 pixels of resolution between the two eyes of the pictured person, and the images released by the FBI on Thursday were about 12 pixels between the two eyes.
Jim Albers, senior vice president of MorphoTrust, a group that supplies facial recognition software to law enforcement officials, says that with a good photo source the software can have just a 0.1% fail rate.
“The limitations are physics”, Albers told Business Insider, explaining that the physical location of the CCTV cameras that caught the images of the Tsarnaev brothers appears to be in a position that created problems (what he referred to as “pose” and “illumination”) with the image. That the suspects were wearing hats, and in the case of the older brother, sunglasses was likely a hindrance too.
Even with advanced “pre-processing” on the images, Albers reasoned that investigators probably only found a low-probability match — not enough to act on.
However, Albers reasons that the inevitable spread of CCTV cameras will likely lead to facial recognition technology being of more use in future cases.
“Cameras will continue to proliferate,” he told Business Insider. “Whether you like it or not, it’s going to happen, and as they continue to proliferate, they are going to get closer to the action.”
NOW WATCH: Briefing videos
Business Insider Emails & Alerts
Site highlights each day to your inbox.