Since 1839, when the daguerreotype camera was first introduced, photographs have documented our history and culture. But it has been challenging for historians to make use of this vast set of visual information using data-mining techniques. Now, computer science Ph.D. student Shiry Ginosar and her team have demonstrated a groundbreaking approach to analyzing large collections of photos. Working with American high school yearbook photographs from 1905 – 2013, the researchers used a dataset of over 37,900 images from more than 800 yearbooks across 26 states. After grouping the pictures by decade and gender, they superimposed the photos to create an average face for each timespan. They also used machine learning techniques to identify distinct clusters of images, as well as deep learning techniques to predict the year in which each image was taken. Their results show the changes over the last century in hairstyles, clothing and even smiles. Previously, researchers would have had to invest significant hours and labor to manually analyze the photos, but the techniques pioneered by the team required minimal human effort. The scientists believe these methodologies can radically change the way in which visual media are used in future humanities research.