The Archival Facial Identification Database is a new way to browse photo collections. By linking similar faces, it highlights social connections that would be difficult or impossible to draw out manually. With the right metadata, it can even help put names to unidentified subjects.
This is a project to bring facial recognition to archival photo collections. The technology has grown by leaps and bounds to meet the parameters of social media and a few other applications, but I believe it can also reveal important historical data, too. There are deep layers of associative meaning in photo collections, which could be invaluable to researchers.
You can read a short process paper about the project here.
The software is built on the Openface project, which uses OpenCV2, Torch, and DLib (among other great open-source software) to detect and classify faces through deep neural networks.