This paper describes a novel interactive clustering system that allows a user to iteratively update a clustering structure. The system uses a 2D projection, both as support for overlaying the current clusters, and to engage user interaction. By clicking elements directly in the visualization, and applying a sequence of operators, the user can inject his domain knowledge progressively, crafting a representation of the data that combines his preferences and the manifolds underlying the data. The interest of the system is demonstrated by detailed experimental scenarios on two image databases.
