Benchmarks for storage and retrieval in multimedia databases

摘要: There is a substantial body of research on computer methods for managing collections images and videos. little evidence that this has had important impact an any community yet. I use invitation to speak topic which am not expert air some opinions about evaluating image retrieval research. In my opinion, there be gained in measuring current solutions with reference collections, because these differ so widely from user needs the exercise becomes empty. The studies literature well enough read by community. As result, we tend study somewhat artificial problems. A suggests will need solve deep problems produce useful problems, but may number technologies can built practice. believe should concentrate issues, rather than performance systems.