The Challenge Problem for Automated Detection of 101 Semantic Concepts in Multimedia
Challenge problem for generic video indexing enables researchers to have a framework for consistent analysis. Due to the ever increasing capture, storage and transmission capabilities, multimedia assets the field of multimedia indexing has experienced a sudden growth over the past decade. Certain problems such as diverse gait analysis face recognition and object detection has provided a baseline for researchers to investigate the issues that interfere with performance. Consequently TREC Video Retrieval is a benchmark that supports the progress in content based access digital video archives through open, metric based evaluation through a common large set of data
There are a number of challenges associated with decomposing the generic video indexing problem into visual only, textual-only, late fusion and combined analysis experiment. However, the general challenge for oftenly appearing ideas is to improve the average precision performance and allow practical utility of video indexing technology in application where efficient performance is needed. Contrary to the semantically linked ideas such as female and monologue still have poor performance.
The challenge problem enhances component based optimization of the generic indexing methodology development. Therefore, it enables insight in factors that influence multimedia evaluation methods, while in the same situation promoting recurrence of experiments. By replacing one or more components, other multimedia indexing researchers may use challenge problem as a foundation of implementation. Consequently, the baseline concept detection is a valuable element for inter active video access experiments. The presence of challenge problem will highly enhance the reliable analysis of generic multimedia indexing algorithms and facilitate research in the multimedia indexing doctrine to weigh against their algorism. Conversely, the challenge of the research reduces the threshold of researchers from other field to engage the disciple of multimedia analysis. (Snoek et al).
Reference
Snoek et al. The Challenge Problem for Automated Detection of 101 Semantic Concepts in Multimedia. Informatics Institute, University of Amsterdam.
