This paper describes partially the architecture and the salient characteristics of a larger visual information retrieval (IR) project for Humanities purposes. The system uses a novel form of interaction for retrieval, sustainable, interactive exploration of retrieval sets, and the types of relationships that are germane to the use of the images, beyond what IR systems typically supply. There are gaps between the expression of a query and the retrieval set and the end-user's ability to establish significance for the set members. The paper sketches the architecture, interaction, and collection metadata created in response to a survey of art history student and instructors' educational needs.
It is argued that a 3D model provides visually-oriented people with a way to comprehend large datasets and, through the interactive control panel and additional graphics, identify facets of value by experts in the field and for local needs.