 |
 |
 |
 |
 |
 |
 |
Workshop
Knowledge in Multimedia Content
November 22, 2006, 11:00-12:30 in Room 206,
Presentations
- Knowledge in Multimedia Content,
Yiannis Kompatsiaris, CERTH-ITI
Yannis Avrithis, IVML-NTUA - Semantic Multimedia Content Analysis,
Yannis Avrithis, IVML-NTUA - Analytic Techniques (Machine Learning & Data Mining) for Semantic Web,
Marko Grobelnik & Dunja Mladenic,
Jozef Stefan Institute Ljubljana, Slovenia - Commercial exploitation of knowledge technologies in communication applications,
Paola Hobson, Motorola Labs - Multimedia Content Analysis for Consumer Electronics,
Jan Nesvadba, Philips Research Eindhoven, The Netherlands - Knowledge representation Products & Services,
Hans-Peter Schnurr, Ontoprise - Semantics in Multimedia Content for Multimedia Use,
Steffen Staab, University of Koblenz - Landau, Germany
Minutes
After the presentations, an open discussion took place and the following topics were discussed:
Multimedia ontologies can be used for sharing multimedia content including a different kinds of associated information and metdata. This results to a number of different type of ontologies including: generic, domain specific, application specific, text and linguistic. There is a need for mapping between these ontologies. Additional information like digital right management has also to be represented. Lessons can be learnt from the text area where such problems have been addressed.
TV producers also have a need for representation of different kind of information. This includes: lighting conditions, size and quality properties for example.
Security applications and retrieval has additional requirements like need for fast decisions which affects the complexity of the ontologies, behavior representation and compliance to standards.
A discussion on what exactly "analysis of broader domain leads to less accurate results" followed. Generic methods could be applied to specific domains for better results. "Generic" analysis modules could be combined in order to be applied to specific domains.
Finally there was a long discussion on evaluation and benchmarking using common datasets in Europe. The possibility for an EC funded initiative was discussed, the relation with the US oriented TRECVID and the issues with annotation, size of datasets and legal issues. There was a comment that BBC will make soon its archive available.
Photos
