Semi-Automatic Content Annotation
Publication from Digital
Media Production, Delivery and Interaction for Platform Independent Systems: Format-Agnostic Media , 1/2014
The requirements on the semi-automatic annotation tools are strongly driven by the need for real-time operation. The result of the annotation should provide mid- or high-level descriptions of the content that are semantically useful for making shot framing and selection decisions. This chapter discusses the requirements related to metadata, and the different types of metadata that are relevant in the format-agnostic production process. The chapter talks about the representation and storage of metadata from audio-visual content annotation. The discussion includes models for the representation of metadata, architectures for automatic content analysis and metadata storage, in particular, for providing time-based indexing and access of content annotations. The chapter considers two applications using saliency information that are relevant to format-agnostic production. The chapter explains approaches that support person detection and tracking in a static camera environment. An approach for assisting automatic tools in the production process is described.