Task-based benchmarking and parameterisation framework for media content analysis modules
Media analysis tools are commonly benchmarked separately, evaluating results of a single analysis step against a ground truth. However, in the context of a system put together for a specific task, certain performance characteristics of the analysis tool might be more important than others, and certain types of errors more critical. The aim is to implement a framework for benchmarking in a system context, based on formalised task models, and to optimise the parameterisation of content analysis modules for a certain system configuration.
Required: C/C++, Perl/Python useful, image/video processing basics usefull.