Can We Exploit Human Visual Attention to Improve Face Recognition Models?

State of the art face detection and recognition algorithms are increasingly focusing on attention, but how do they compare with human attention on faces? Do they “fixate” the same areas as humans do? Can we use human fixations to improve recognition models? We try to answer these questions using a human fixations dataset on face images, comparing them to Visual Transformers and using them to build a bottom-up, part-based model for face detection and recognition.

Curriculum vitæ

Marinella Cadoni received her PhD on Mathematics at the University of Warwick (UK) in 2004. From 2004 to 2008 she worked as a postdoc at the IEIIT-CNR research institute located in Turin, where she carried out research in the field of Computer Vision. She then won a three year research grant on Biometrics at the University of Sassari. From 2011 to 2017 she carried out research activities at the University of Sassari and for private companies. From 2017 she is an Assistant Professor in Computer Science at the University of Sassari. Her research interests focus on Biometrics, 3D object reconstruction and retrieval and human visual attention.

We thank our media partner: