Skip to main content

Video indexing classification annotation and retrieval

Objective

The overall aim of the project is to develop a system which includes each of the functions required to index & search videos. The primary focus is the professional video archive market, for which a pilot system will be developed. The requirements for adapting the software to smaller scale operations (consumer market) will be taken into account. The project will install a pilot version at the end user sites to measure user acceptance. The end users will create indexes of a high percentage of their digital video material and provide them to a central site, which will maintain them on an HPC system and provide a global search service. A WWW interface will enable access to the system. Quality and robustness of the search results is a major technical goal.

The project will develop a software system for the automatic generation of semantic indexes for video and for associative search in the index structures thus formed. This environment will function as a learning system that is trained on reference material (video with content descriptions) to construct relevant high dimensional feature representations.

The product Video Explorer of the project is a semantic indexing and retrieval engine for video databases. The project will develop software products for the creation of video indexes and for searching within such an index. It will also build up a prototype service which makes use of this software and the parallel computer at VECPC to enable a wider-scale search and retrieval of centrally stored index material. Although the service will begin with content provided primarily by the end users in this project, it is believed that in the longer term there is a large potential market both on the supply as well as on the demand side. A new company will be formed in order to fully exploit the results of the project.

Funding Scheme

CSC - Cost-sharing contracts

Coordinator

JOANNEUM RESEARCH FORSCHUNGSGESELLSCHAFT MBH
Address
Steyrergasse 17
Graz
Austria

Participants (2)

Sentient Machine Research
Netherlands
Address
Baarsjesweg 224
1058 AA Amsterdam
Österreichischer Rundfunk
Austria
Address
Würzburggasse 30
1130 Vienne