1. The CUbRIK project combines automatic face recognition tools with crowdsourcing to build social graphs from images. It uses a four-tier architecture to process content and queries.
2. An initial proof of concept used a general purpose crowd to validate top face matches identified by an automatic tool on a dataset of European history photos, but achieved only weak results due to image quality issues and the non-expert crowd.
3. The presentation argues that using an expert-based crowdsourcing approach could overcome the limitations seen with solely automatic processing and general purpose crowds, by leveraging experts' domain knowledge and ability to collaborate within their communities.