Who's Behind the Camera? Identifying the Authorship of a Photograph

We introduce the novel problem of identifying the photographer behind the photograph. To explore the feasibility of current computer vision techniques to address the problem, we created a new publicly downloadable dataset of over 100,000 images taken by 25 well-known photographers. Using this dataset, we examined the effectiveness of a variety of features (low and high-level, including CNN features) at classifying the photographs. Our quantitative and qualitative results illustrate that high-level features significantly outperform low-level features at this task. We also provide additional qualitative results demonstrating how these learned models can be used to draw interesting conclusions about what specific photographers tend to shoot.