March 12 2008, 14.00 hrs

Robin Aly

Abstract:

The major challenge in multimedia information retrieval is the semantic gap. To tackle this problem, the research community came up with the notion of semantic concepts, which are bundling abstract low-level features to features which have meaning to users: outdoor, car, people. As classification techniques (concept presence/absence) are not precise enough, detector return probabilistic output. Therefore, the search systems have to deal with the uncertain occurrence of a feature. It is difficult to combine multiple uncertain concepts to a score for the document. This talk describes these problems more in-depth and proposes a method to exploit timely connection of video material and a method for pure concept search.

	Home > ... > Past colloquia > March 12 2008, 14.00 hrs
<< EWI Home Staff Research Teaching Publications Intranet Colloquia DB Vacancies News Search	March 12 2008, 14.00 hrs Robin Aly Abstract: The major challenge in multimedia information retrieval is the semantic gap. To tackle this problem, the research community came up with the notion of semantic concepts, which are bundling abstract low-level features to features which have meaning to users: outdoor, car, people. As classification techniques (concept presence/absence) are not precise enough, detector return probabilistic output. Therefore, the search systems have to deal with the uncertain occurrence of a feature. It is difficult to combine multiple uncertain concepts to a score for the document. This talk describes these problems more in-depth and proposes a method to exploit timely connection of video material and a method for pure concept search.