IDENTITY CRISIS: WHEN FACE RECOGNITION MEETS TWINS AND PRIVACY

Abstract

Face recognition has received extensive attention during the past several decades. Most works focus on improving recognition accuracy for general population, while limited works have studied identical twins and privacy protection. For identical twins, their appearances are quite similar, thus conventional face recognition systems utilizing appearance may be not suitable anymore. For privacy protection, most existing works protect privacy at the expense of destroying utility of other facial attributes, such as gender and race. These attributes, however, may be beneficial for many visual analysis systems. In this thesis, we try to address these two problems by 1) utilizing behavioral biometrics and ear biometric to recognize twins 2) proposing a new scheme, SemanticFace, which can independently control the amount of identity alteration while keeping other facial attributes unchanged to protect privacy without thwarting facial analysis systems. The extensive experiments show the effectiveness of our proposal on twin recognition and privacy protection.