Computer Science: Bombs Away!

We all know the feeling of composing the perfect picture and then, just as you push the button, a passing car enters the frame — or your hilarious cousin Joe photo-bombs the shot. Ohad Fried, a graduate student in computer science, is developing an app that automatically will remove intrusions from a photo after it is taken.

Ohad Fried GS

Courtesy Ohad Fried GS and Adam Finkelstein, Department of Computer Science, Princeton University; and Eli Shechtman and Dan Goldman, Adobe Research

Fried, who is working with computer science professor Adam Finkelstein on the project, pooled data in which “distractors” had been removed in order to create an algorithm that identified elements such as cars, signs, trash on the ground, spots on the lens, and, yes, faces coming in from the side of the picture that might represent a photo-bomb. He trained the computer to recognize those patterns and fill in the spaces with a composite of the surrounding area. Results so far haven’t been perfect — in some cases, vital elements have been removed, or holes have been filled in with equally distracting blobs. But Fried is working on refining the program, which he is creating with software company Adobe.