GRIM Management and summarization of multimedia data
Management and summarization of multimedia data
The GRIM team deals with questions arising from the search for information in large volumes of data (relational, multimedia).
The central point is the creation and maintenance of parsimonious structures of data, based in particular on fuzzy sets or parametered probabilistic models. Obtaining these data structure involves unsupervised classification tasks for the main part.
The structure obtained can, depending on the case, allow effective (in "system" terms) and flexible (managing uncertainty and inaccuracy) querying of large amounts of data, or/and provide interpretable classes and interactive/navigation possibilities.
We also study construction and maintenance in cases where the data is dynamic: data flows, distributed data (or to be distributed): peer to peer, cluster.