Privacy-Aware Methodologies For Person Re-Identification

Person Re-Identification (Re-ID) is a computer vision task that involves retrieving target individuals across non-overlapping cameras. Despite its utility in intelligent surveillance systems, its deployment within the European Union poses notable challenges in terms of legal compliance. This inspired us to investigate the particular requirements for developing a compliant and performing Re-ID system. To this end, we propose a compliance framework, derived from European regulations and guidelines, to evaluate privacy-preserving solutions. We apply this methodology on a selection of case studies to highlight the different strategies and trade-offs in creating such techniques, as well as offering guidelines to lawfully design Re-ID systems.