URJCyUNED at ImageCLEF 2012 Photo Annotation Task.

This paper describes the URJCyUNED participation in the ImageCLEF 2012 Photo Annotation task. The proposed approach uses both visual image features and textual associated image information. The visual features are extracted after preprocessing the images, and the textual information are the provided Flickr user tags. The visual fea- tures describe the images in terms of color and interesting points, and the textual features make use of the semantic distance between the user tags and the concepts to annotate by using WordNet. The annotations are predicted by SVM classiers, in some cases trained separately for each concept. The experimental results show that the best of our submissions is obtained by using only textual features.