For this project, I chose to explore the ROAD 2.0 library of outdoor advertisements via Duke University Libraries.
Because advertising reflects target demographics, and advertising en masse a supposed ideal of society, I was drawn to looking at the people depicted in the ads. In order to separate the people from the context of the products being advertised, I ran a script to detect faces and cut them out.
Using the approximate dates in the image metadata, I sequenced the faces in a rough chronological order and created an animation. For this animation I also included the text from the ads; it’s a rough attempt to look at not just trends in facial appearance, but the overall tone of the messaging, which still trying to abstract from the products and companies.
Face detection code (Open Framworks, includes Python image/data scraping scripts): via github
Animation code (Open Frameworks): via github
t-SNE grids of subsets of the library (click to enlarge):
