. German Research Center for Artificial Intelligence GmbH
The project VITRA (VIsual TRAnslator) deals with the relationship between natural language and vision. Experimental studies are being carried out in the way of designing an interface between image-understanding and natural language systems, with the aim of developing systems for the natural language description of image sequences.
Together with the vision group at the IITB, Karlsruhe, first results in connecting a vision component and a natural language access system have been obtained. These previous attempts have been restricted to a bird's eye view and thus 2-dimensional representation of the scenes under discussion. Within the new VITRA system the methods developed so far will be extended in order to cope with 3-dimensional visual information.
In this contribution the capabilities of the VITRA workbench concerning 3-dimensional geometric representations will be presented. The paper will concentrate on the visualization of the geometric data as well as on the representation and manipulation of the underlying image sequences.