Taylor Arnold and Lauren Tilton

Computational Approaches for Studying a Large Photographic Corpus

Paper Presentations IIAnalysis and discovery models for audiovisual materials

 When applied to large digitized archives of images, computational and statistical methods offer new approaches for analysis and often lead to novel insights. In this talk we present a case study of this approach from the FSA-OWI Photographic Archive, a collection of over 170,000 photographs taken by the United States Government between 1935 and 1945. We will first show how exploratory and graphical techniques reveal important spatial and temporal patterns in the archive as a whole. We then give several examples of how computational techniques can be used to infer latent metadata about individual records. These examples illustrate how computational techniques are useful not just in aggregate, but also in the close analysis of even a small set of digitized images. We finish by demonstrating how the output from these analyses is being integrated into Photogrammar, our web-based platform for exploring the FSA-OWI collection.