Presenter(s): Alberto Messina (RAI), Alexandre Rouxel (EBU), Sébastien Ducret (RTS)
Event: MDN 2022 Online Conference
This presentation reviews the key technologies and metrics used by an open source framework for facial recognition developed by the EBU AI Benchmarking working group. The developed pipeline will bring together models and algorithms for facial recognition on video. The aim is to have a parametric pipeline to evaluate different face detection and recognition options, as well as other blocks such as clustering. The tool will generate scores by considering metrics well suited to videos.