Editors’ Choice: Using Computer Vision to Increase the Research Potential of Photo Archives

The application of computer vision to art photo archives has largely been unexplored up to this point. Lev Manovich has explored ways of analyzing images of artworks while looking for trends in an artists oeuvre or entire artistic movements. However, most institutions have used large scale image analysis primarily for cases of copyright enforcement, face detection, or color/composition analysis.

To explore what image similarity analysis was capable of, I completed an analysis of the digital images of Italian anonymous art at the Frick Photoarchive. The image similarity analysis, using TinEye’s MatchEngine service, was automated using newly-developed tools. I further processed and dissected the data using custom tools. The analysis was able to confirm some of the existing relationships between photographs that were manually generated by researchers. The analysis was also able to discover a number of completely new relationships, including: works of art before and after conservation, copies of the same artwork, cropped detail shots of the same artwork, and cataloging errors.

Read Full Post Here

This content was selected for Digital Humanities Now by Editor-in-Chief Amanda Morton based on nominations by Editors-at-Large: Caitlin Christian-Lamb, C├ęsar Viana Teixeira, Ester Rincon Calero, Beth Secrist, Amy Williams, Dale Russell, Aisha Clarke, Silvia Stoyanova, and Melanie Baptista