How to visually retrieve images from the St. Andrews collection using GIFT

Abstract

The ImageCLEF task of CLEF has a main goal in the retrieval of images from multi-lingual collections. The 2003 imageCLEF saw no group using the visual information of images, which is inherently language independent. The query topics of the St. Andrews collection are defined in a way that makes visual retrieval hard as visual similarity plays a marginal role whereas semantics and background knowledge are extremely important, which can only be obtained from text. This article describes the submission of an entirely visual result. It also proposes improvements for visual retrieval systems with the current data. Section 4 explains possible ways to make this query task more appealing to visual retrieval research groups, explaining problems of visual retrieval and what the task can do to overcome present problems. A benchmarking event is needed for visual information retrieval to remove barriers in performance. ImageCLEF can be this event and identify areas where visual retrieval might be better than textual and vice-versa. The combination of visual and textual features is an important field where research is needed. © Springer-Verlag Berlin Heidelberg 2005.