Agent WiSARD in a 3D World

摘要

This paper investigates the integration of verbal and visual information for describing (explaining) the content of images formed by three-dimensional geometrical figures, from a hybrid neurosymbolic perspective. The results of visual object classifications involving top-down application of stored knowledge and bottom-up image processing are effectively explained relying on both words and pictures. The latter seems particularly suitable in explanations concerning high-level visual tasks involving both top-down reasoning and bottom-up perceptual processes.