In this demonstration we present SHIATSU, an automatic semantic-based video tagging system which relies on shot boundary detection and hierarchical annotation. More in details, in SHIATSU, shots obtained from video segmentation are first automatically labelled and such labels are then propagated at the video level. The approach is novel and appealing because: 1) it only considers the visual content of each video in order to automatically suggest description labels, 2) predicted tags can be

more »

... wed/accepted by the user and persistently stored in the system in order to be exploited when searching for video of interest, and 3) the provided GUI makes annotation of videos extremely intuitive and usable.