Kyle Banerjee and Maija Anderson have published "Batch Metadata Assignment to Archival Photograph Collections Using Facial Recognition Software" in the latest issue of Code4Lib Journal.
Here's an excerpt:
Useful metadata is essential to giving individual meaning and value within the context of a greater image collection as well as making them more discoverable. However, often little information is available about the photos themselves, so adding consistent metadata to large collections of digital and digitized photographs is a time consuming process requiring highly experienced staff.
By using facial recognition software, staff can identify individuals more quickly and reliably. Knowledge of individuals in photos helps staff determine when and where photos are taken and also improves understanding of the subject matter.
This article demonstrates simple techniques for using facial recognition software and command line tools to assign, modify, and read metadata for large archival photograph collections.