DIVA - Digital Video Analysis

Michel Beaudouin-Lafon, mbl@lri.fr

Wendy E. Mackay, mackay@lri.fr

DIVA est un système pour l'analyse de données vidéo grâce à une algèbre de streams qui permet de calculer sur des données temporelles et différents types de visualisation interactive de ces données.


Wendy E. Mackay and Michel Beaudouin-Lafon
DIVA: Exploratory Vido Analysis with Multimedia Streams
Proc. ACM Human Factors in Computing Systems (CHI'98), Los Angeles (CA), ACM Press, New York, pp 416-423.


DIVA supports exploratory data analysis of multimedia streams, enabling users to visualize, explore and evaluate patterns in data that change over time. The underlying stream algebra provides the mathematical basis for operating on diverse kinds of streams. The streamer visualization technique provides a smooth transition between spatial and temporal views of the data. Mapping source and presentation streams into a two-dimensional space provides users with a direct manipulation, non-temporal interface for viewing and editing streams.

DIVA was developed to help us analyze both qualitative and quantitative data collected in our research with French air traffic controllers, including video of controllers at work, audio records of telephone, radio and other conversations, output from tools such as RADAR, and coded logs based on our observations. Although our emphasis is on exploratory data analysis, DIVA's stream architecture should prove useful for a wide variety of multimedia applications.

Images et transparents

Quelques images de DIVA sont disponibles, ainsi que les transparents de la présentation de l'article ci-dessus à la conférence ACM CHI'98.

La présentation des transparents nécessite un navigateur qui implémente le langage JavaScript.


Michel Beaudouin-Lafon, mbl@lri.fr, http://www-ihm.lri.fr/~mbl