-
Framework:
- Start: 28/02/2007
- End: 30/09/2008
- External site
This project has concluded.
In the context of the research project PISA we have investigated the production, indexing and search of Audiovisual material.
The media industry, and in particular the production of radio, television and feature film, is characterised by a number of typical production processes. As applicable in any other industry, these are preceded by a phase of design and development referred to as Product Engineering. While this process is well known, structured and formatted, the results as such are usually a textual and unstructured representation of the programme or product. Then again, a script can only contain a relative small fraction of the creative decisions that are required to create a finished product. This situation leads to trial-and-error during production and in general the reproducibility of the production process is sub-optimal.
In PISA we have investigated how the product engineering process can be optimised by model-driven development, i.e. by using techniques of Computer Assisted Manufacturing (CAD/CAM). We have been able to demonstrate how the productivity is increased by a more structured interpretation of product engineering decisions. In particular we have been able to automate various ingest processes, the rough-cut editing and the reproduction of audiovisual products as function of the display format.
In addition we have investigated the inverse process, referred to as archiving. We have developed various techniques for automatic and semi-automatic image analysis and we have investigated to which extent and under which conditions they were able to provide value added compared to conventional techniques and search engines. We have built a powerful search engine which is designed to disclose audiovisual material and which is specifically optimised to operate in an enterprise context, characterised by finite number of sources and whereby production processes and use cases provide prior knowledge during indexing. We have been able to dramatically increasethe precision and the recall of the extracted information and by extension to increase proportionally the efficiency of the overall search experience.