A quest for human identity recognition - problems and methods of facial re-identification
Damian Pęszor
Polish-Japanese Academy of Information Technology and Silesian University of Technology
Abstract
Among the many biometric features that can allow the recognition of a person's identity, most are associated with invasive methods of data acquisition that require temporary involvement and awareness of the recognized person. In many circumstances, for example, in screening for wanted persons, such biometric features are not applicable. Humans, social creatures, recognize identity primarily based on a non-invasive method, using the human visual system, the end of which is a complex process of facial image analysis evolving along with social and cultural abilities.
The face re-identification process, outside of laboratory conditions, is burdened with many difficulties related to the variety of acquisition conditions and the importance of details of the human appearance. As part of the talk, the stages of the facial re-identification pipeline will be discussed along with the problems that are addressed by these stages, their variants and the impact on the complex process of facial re-identification.
About the Author
Damian Pęszor received B.Sc and Ph.D. degrees in computer science from the Polish-Japanese Academy of Information Technology (PJAIT), Poland, in 2011 and 2018, and M.Sc. from the Silesian University of Technology (SUT), Poland, in 2012, wherein he finished his doctoral studies in 2017. He is currently an assistant professor at the Department of Computer Graphics, Vision and Digital Systems of SUT and a researcher and head of programming at the PJAIT Research and Development Center. His current research focuses on computer vision, computer graphics, machine learning, hierarchical clustering, computer-aided diagnosis and motion analysis. He was also involved in research projects related to facial animation and reidentification, quantum computing, virtual reality and autonomous vehicles.