UBEAR: A dataset of ear images captured on-the-move in uncontrolled conditions

Abstract

In order to broad the applicability of biometric systems, the data acquisition constraints required for reliable recognition are receiving increasing attention. For some of the traits (e.g., face and iris) significant research efforts were already made toward the development of systems able to operate in completely unconstrained conditions. For other traits (e.g., the ear) no similar efforts are known. The main purpose of this paper is to announce the availability of a new data set of ear images, which main distinguishing feature is that its images were acquired from on-the-move subjects, under varying lighting conditions and without demanding to subjects any particular care regarding ear occlusions and poses. The data set is freely available to the research community and should constitute a valuable tool in assessing the possibility of performing reliable ear biometric recognition in such d challenging conditions.