When Will Facial Recognition Work Through Halloween Masks?
Spoiler: probably not for a while.
Halloween is around the corner, and with it, a new wave of twenty-somethings who really should know better are set to don an array of scary masks. Although surveillance has reached a stage where the FBI can track your face without you knowing, it seems the state’s ultra-powerful snooper cameras are still bamboozled by the hordes of disguised people walking around on October 31. That, it seems, isn’t going to change soon.
Covering your face is still an effective way of hiding your identity, and though it may seem like a problem that should have been solved in the age of high-tech artificial intelligence, research shows there’s a long way to go. In a paper published in IEEE Xplore in May, Ajay Kumar and Tsung Ying Wang from the Hong Kong Polytechnic University found that two commercial products promising to detect faces through disguises fell far short of the mark. The team then produced a public database aimed at helping facial recognition software understand what parts of a person’s face are under a disguise, allowing it to “see” the real face, and come to a decision.
“The experimental results underline the challenges in recognizing faces under these covariates,” the abstract reads. “Availability of this new database in public domain will help to advance much needed research and development in recognizing makeup and disguised faces.”
Creepy, and also hard to detect.
Another technology that could help cameras see past disguises is infrared recognition. That may help cameras see past masks, but nothing’s been promised yet.
“I’ve spoken with engineers at a couple of companies who are in the early stages of exploring infrared, but I’m not at liberty to share whom specifically,” Jonathan Frankle, staff technologist at the Center on Privacy & Technology at Georgetown Law, tells Inverse. “In an ideal world, the technology might avoid issues of obstruction, lighting, and even skin tone (hence, racial bias), but there’s currently no public data to support these hunches.”
It may seem like an odd fixation, but with the spate of creepy clown attacks across the world, improvements in recognition technology could help police catch the perpetrators in situations like these.
“As a technologist, experience has taught me better than to ever try to estimate how long technology will take to move from idea to product,” Frankle said. “A lot of technologies fail along the way, and it’s very possible infrared FR would suffer that fate. Nuclear fusion has been 20 years away for what, 70 years now?”
