September 7, 2017 -
Researchers are developing an AI-powered facial recognition system designed to help identify criminals, political dissidents, protesters, and other individuals who attempt to disguise their identity by hiding their face with masks, hats, sunglasses, or other accessories, according to a report by International Business Times.
Disguised Face Identification (DFI) leverages a deep-learning neural network trained on a dataset of images of people using different accessories to conceal their faces with various backgrounds.
The technology maps a total of 14 facial points (10 for eyes, one for the nose, and three for lips) on an individual’s face and measures the distance and angles between those points to estimate the concealed facial structure.
The system then compares the approximated facial structure with learned images to help identify the person in question.
Based on preliminary tests, the deep-learning algorithm delivered 56 percent identification accuracy when the face was concealed with hats or scarves. When glasses were added to the test subject, the accuracy level dropped down to 43 percent.
The research team — comprised of members from the University of Cambridge, National Institute of Technology, and Indian Institute of Science – understand that the AI-based facial-recognition system is still at an early stage and will require several improvements before being applied in the field.
As a result, the team has released datasets of disguised and undisguised faces, and have asked other experts to test and develop the technology.
The research offers insight into the potential applications of a technology that could help identify people by simply scanning their concealed faces, with law enforcement being an obvious field that could benefit from the technology. However, the application could raise concerns for violating the privacy of several people who regularly wear hats and scarves.
“This is very interesting for law enforcement and other organisations that want to capture criminals,” Amarjot Singh, a researcher at the University of Cambridge who worked on DIF, told Inverse. “The potential applications are beyond imagination.”
The research team is currently preparing to present the research at the IEEE International Conference on Computer Vision Workshop, held in Venice, Italy from October 22 to 29.