Which tags are related to visual content?

Abstract

Photo sharing services allow user to share one's photos on the Web, as well as to annotate the photos with tags. Such web sites currently cumulate large volume of images and abundant tags. These resources have brought forth a lot of new research topics. In this paper, we propose to automatically identify which tags are related to the content of images, i.e. which tags are content-related. A data-driven method is developed to investigate the relatedness between a tag and the image visual content. We conduct extensive experiments over a dataset of 149,915 Flickr images. The experimental results demonstrate the effectiveness of our method. © 2010 Springer-Verlag Berlin Heidelberg.