Detecting People in Artwork with CNNs

TL;DR

This paper demonstrates that fine-tuned CNNs can effectively detect people across diverse artwork styles, achieving state-of-the-art results on the challenging People-Art dataset, though further improvements are needed.

Contribution

It introduces a CNN-based approach for detecting people in artwork and shows the importance of transfer learning limitations between photos and artwork.

Findings

Achieved over 60% AP on People-Art dataset

Fine-tuning CNNs improves detection in artwork

Only initial CNN layers transfer well from photos to artwork

Abstract

CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset, People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by fine-tuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN's performance is the highest yet, it remains less than 60\% AP, suggesting further work is needed for the cross-depiction problem. The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-319-46604-0_57