Thanks to easy-to-use multimedia devices and cheap storage and bandwidth, present-day social media applications host staggering numbers of personal photos. As the number of personal photos shared on social media applications continues to accelerate, the problem of organizing and retrieving relevant photos becomes more apparent for consumers. Automatic face recognition assists in bringing order to collections of personal photos. However, personal photos pose a plethora of challenges for automatic face recognition. Face images may widely differ in terms of lighting, expressions, and pose. As a result, the accuracy of appearance-based techniques for automatic face recognition in collections of personal photos cannot be considered satisfactory.
This talk aims at providing insight into timely developments in the area of socially-aware face recognition. We first discuss how online social network context can be used to substantially improve the effectiveness of appearance-based techniques for automatic face recognition, as recently demonstrated by researchers of Harvard University. Next, we pay attention to collaborative face recognition in decentralized online social networks, as studied at KAIST. For both of the aforementioned topics, we present experimental results obtained for real-world collections of personal photos, contributed by volunteers who are members of online social networks such as Facebook and Cyworld. Finally, we conclude our talk with an outline of future applications of socially-aware face recognition, including augmented identity and socially-aware robots.