You can watch the presentation videos of some prototypes of semantically enhancend image search engines.

• ThemExplorer A demonstration that mashes up TagMaps (maps service offered by Yahoo!), Gazetiki(a world-scale geographical gazetteer) and a content-based image retrieval system over pictures from Flickr and Google Images. The user can select one or more themes of interest (from a set of around 100) and see images of geographic entities related to those themes in an area delimited by the current view of the map. The content-based search includes only the images that belong to the same query, ensuring both a conceptual and a perceptual coherence of the answers.
• Olive is a semantic layer over existing Web image search engines. With the introduction of the ontology in the retrieval framework, the system automatically reformulates user queries to propose pictures for the subtypes of the demanded concepts. This approach results in an improvement of the obtained results and extended interaction possibilities. The system employs a conceptual hierarchy including over 117000 English terms (extracted from WordNet) and Google images that are collected on the fly.
• SemRetriev an earlier demonstration using the same functionning principle as Olive, but working with an ontoology of 1113 classes (placental animals) and a local picture repository of around 25000 items (collected with Ask) attached to the leaves of the hierarchy. The utilization of a local structured database allows the proposition of conceptually controlled CBIR processes with different generality degrees.