A joint US research Project from the University of Illinois at Urbana-Champaign and Disney Research, Pittsburg, has made steps towards an automated digital image family member detection system. The new Bayesian photo-analysis methodology is able, for example, to perform ”identity clustering” and make attempts at identifying not only which individuals might be mothers, fathers, and children – but also grandparents.
“[…] family photo collections have unique challenges and opportunities for face recognition compared to random groups of photos containing people. We address the problem of unsupervised family member discovery: given a collection of family photos, we infer the size of the family, as well as the visual appearance and social role of each family member.”
Future work is expected to expand the inferred family connections to aunts and uncles.
“Interesting directions for future work is to add roles for non-family members and additional relatives, such as aunts and uncles, and to consider other types of social groups, such as friends”