Invited Speaker

Marco Huber

 Towards Trustworthy Face Recognition

Modern face recognition solutions rely heavily on deep learning models whose internal logic and decision-making processes remain hidden from the user. As these biometric systems are increasingly deployed in high-risk scenarios, such as automated border controls, and process highly sensitive, private data, the need for transparency, bias mitigation, and trustworthiness has become crucial. This talk covers recent advancements toward achieving trustworthy face recognition by exploring three key pillars: uncertainty & estimation, explainable face recognition in the spatial domain, and explainable face recognition in the frequency domain.