Interactive Face Recognition System
Our innovative Interactive Face Recognition System (I-FRS) uses naturalistic 3D models of faces and allows the user to dynamically view faces from multiple angles.
We work with academics from all over the world to conduct research using the I-FRS. Research shows that the I-FRS can improve human face recognition and face-matching accuracy, compared to other methods currently used by criminal justice, border control, and law enforcement agencies. The I-FRS can also help us to understand how humans process and identify faces.
We work closely with both criminal justice professionals and technology agencies to ensure that the I-FRS is regulatory compliant, technically proficient, and an effective system that can solve real world face recognition problems.
Can interactive Lineups Reduce Own Race Bias?
People are less able to accurately perceive and remember faces that are of a race different to their own, a phenomenon known as own race bias (ORB). The causal mechanisms behind ORB have been attributed to both cognitive and social processes at encoding. To date, ORB has been predominately tested using static photographs of people shown head on, in frontal pose. Hence, it is unknown whether the ability to actively explore a face to view it in other poses would enhance memory retrieval. We are using an eyewitness memory paradigm to test whether an interactive lineup procedure that allows participant witnesses to rotate and view the lineup faces in any pose (e.g., profile, ¾ view) increases discrimination accuracy, especially for cross-race witnesses. The results are presented in the slideshow below and a pre-print of the paper is available here.
This work was generously funded by the Laura and John Arnold Foundation
Can pose reinstatement improve eyewitness identification accuracy?
Police around the world ask witnesses to identify criminal suspects from lineups composed of static photographs showing each lineup member facing the camera, head-on, in frontal pose. Yet, for witnesses who encoded the perpetrator in other poses (e.g., ¾ or profile view), encoding specificity predicts identification accuracy will be higher if they could reinstate perpetrator pose, and view the lineup faces in the same angle as they had encoded the perpetrator. We are testing the effects of pose reinstatement on identification accuracy, and whether witnesses would actively seek to reinstate perpetrator pose at test, if given the opportunity. The results from one of our studies is given below in the slideshow, and a pre-print of one of our papers is available here.