Face-recognition systems tested on identical twins
25 March, 2011
category: Biometrics
Researchers from the University of Notre Dame, Kevin Bowyer and Patrick Flynn, have taken to the Twins Days Festival in Twinsburg, Ohio to collect participants to assist them in testing the limits of face recognition technology, according to an IEEE Spectrum article.
The research performed by the two professors was funded by a grant from the FBI as they have had a long-standing interest in face recognition technology both in operation as well as seeing it improve enough to be usable as evidence in court proceedings.
While the testing was performed at the twins festival in 2009, Bowyer and Flynn presented their findings at the IEEE International Conference on Automatic Face and Gesture Recognition 2011 revealing that face recognition systems are not yet capable of perfectly distinguishing between two with such similar features, but show promise in certain situations and environments.
Environments wherein the face recognition systems tested very well at telling differences between identical twins include controlled ideal lighting such as studio lights and controlling facial features such as smiling.
However, when the researchers performed the same tests under conditions that more closely resembled real-world conditions the systems were unable to tell the differences well at all. In order to rectify this moving forward, Bowyer and Flynn suggest a push for higher resolution cameras focusing on finer facial details to get a truly usable and effective facial recognition system developed.
Read the full story here.