Towards robust and reliable multimedia analysis through semantic integration of services

摘要: Thanks to ubiquitous Web connectivity and portable multimedia devices, it has never been so easy produce distribute new resources such as videos, photos, audio. This ever-increasing production leads an information overload for consumers, which calls efficient retrieval techniques. Multimedia can be efficiently retrieved using their metadata, but the analysis methods that automatically generate this metadata are currently not reliable enough highly diverse content. A automatic method analyzing general content is needed. We introduce a domain-agnostic framework annotates available methods. By three-step reasoning cycle, assess improve quality of results, by consecutively (1) combining results effectively, (2) predicting might need improvement, (3) invoking compatible retrieve results. semantic descriptions services wrap methods, selected. additional on these descriptions, different repurposed across use cases. evaluated problem-agnostic in context video face detection, showed capable providing best regardless input video. The proposed methodology serve basis build generic annotation platform, returns problems. allows better generation, improves resources.