PR: The Average CP2 student

Each year in my Computational Photo II class, I ask students create an average face. We start by exploring artwork and historical experiments, such as Francis Galton’s research into identify trends in faces. Galton used optical averaging to visualize facial similarities and differences, seeking to identify common features among criminals and Jewish people (among others). Artists have also explored image averaging: Jason Salavon is the best known example for his work averaging images of children with Santa and centerfolds from men's magazines of the 1960s and 70s.

This technique has been applied to identify standards of beauty in different cultures by averaging faces from beauty pageants or advertisements to produce a composite face embodying common features. The exercise is culturally rich and serves as a great icebreaker in class.

They set up a camera either in the classroom or outside. Each student stands in a designated spot, and aim for a rough alignment of their faces. Sometimes they make two poses—a neutral expression and a smiling face—or have them remove eyewear to better capture their everyday appearance.

I have several examples from different years, and you can see how the "average" computational photography student has changed. Although these averages are not entirely scientific, they were created by students using their own faces and following my instructions. Through this process, students not only learn about computational photography but also about each other, moving beyond surface impressions.

From a technical perspective, the software used in this exercise isn't particularly difficult, but it does require a facial detection model similar to those used in football stadiums for identifying criminals. These models are widely available, lightweight, and even integrated into modern surveillance cameras. This sets the stage for a later exercise where students try to confound face detection algorithms by wearing disguises or using makeup, helping them understand how these systems work and explore ways to evade detection.