The Art Historical Image Collection at Columbia University: Automating Research on Its Construction and Creators

This paper aims to communicate an ongoing research project conducted at Columbia University’s Media Center for Art History. This on-going project evaluates technologies from artificial intelligence and data science to automate the extraction of image and catalog information from a large image collection held in the department of Art History and Archaeology at Columbia. The project addresses image processing, neural networks, and decision trees to recreate the organization of the collection for which no general catalogue exists and also investigates how the research and publications of the department’s faculty can be linked to the information hidden in the repository of images. The Media Center has successfully used different technologies such as deep learning to study the collection of over 600 000 image assets.

For example, software was configured and deployed to detect original photographic images in the 35mm slide collection. These technologies serve to classify images as copywork or an original photo. The classification image detection workflow developed by the Media Center has applications beyond art history to image databases throughout the humanities, especially if the collection suffers from a lack of metadata. This research has the potential to participate in the creation of large datasets of art historical visual materials, which, in light of the continued progress of new technologies, can lead to news discoveries.

Stefaan Van Liefferinge, Gabriel S. Rodriguez & Lisa Peck (Media Center for Art History, Columbia University)

Projet MemoRekall