When you look at the photo above, what do you see? A construction site? Three people working with cement while another supervises? A line of trees? You might recognize traditional Parisian buildings and so see a story about urban change in the French capital. A computer might only be able to see some of what you do: recognize the trees, say, or count how many people are present. But what if we tell you that there are 100,000 such images, all taken in the city during a single month in 1970? While you could never easily look at every photograph, a computer can. How can we leverage the perspective that computers bring to such a substantial visual dataset with such a quantity of images? How can they help to facilitate rich historical and formal readings not just of a single photo, but images numbering in the tens of thousands?

This semester, the lab is working with Faculty Fellow Catherine Clark to realize the potential of these techniques for the study of photography. Associate Professor of History and French Studies, Clark is a cultural historian of modern Europe who has written widely about the histories of modern France and its visual culture. Her 2018 book, Paris and the Cliché of History, from which her DH project stems, was winner of the 2018-2019 Laurence Wylie Prize in French Cultural Studies. At MIT she teaches classes about the histories of modern Europe and urban culture, the history of photography, the role of Paris as a global capital, and contemporary French culture.

Prof. Clark will be leading our team of UROPs in enhancing our understanding of the "This was Paris in 1970" digital archive—a collection of over 100,000 photographs from a Parisian amateur photography competition of that year—through computer vision and machine learning techniques. The contest “C’était Paris en 1970” was organized by the FNAC and the City of Paris, and its submissions are currently held at the Bibliothèque historique de la Ville de Paris. When the work is complete, the framework will be available to digital archives seeking to achieve similar goals with different visual data sets. Ultimately, the project will help researchers identify and analyze the contents of visual archives: not only will a computer "see" the construction and the people working. It will also lay the groundwork for more nuanced narratives, drawing upon the computer's ability to recognize how many other photos in the archive “look” like this one and how typical of a particular photographer’s submissions a single image may be. Working in this way holds the potential to transform the narratives we tell, as we bring the mutually-informing worlds of computational and human interpretation ever closer.

You can read more about Prof. Clark's work here.