Vast amounts of Multimedia data are being generated nowadays all around the world. A number of complementary factors have contributed to this explosion of content: affordability of high quality capturing gear, miniaturization and ubiquity of devices, and, especially, the Internet as an efficient technology to store and transmit the data. The Internet has deeply impacted the way we interact with information. It has provided the means for every user to air their own contents with extreme ease. The traditional information flow model, coming from a few sources to many consumers, has been replaced with this many-to-many paradigm. A plethora of problems related to search and discovery of resources is generated in this complex scenario: users have difficulties finding the content they care about, and demand facilities to discover new potentially appealing assets. While traditional Information Retrieval research has investigated solutions to search and discover relevant content in large-scale databases of text documents, multimedia information is still difficult to retrieve using automatic tools. Multimedia analysis methods find their main limitation in the so-called semantic gap, i.e. the difficulty of inferring high-level concepts from the set of low-level features which can be extracted from multimedia data. The objective of the MIESON project is to address the shortcomings of state of the art technology in multimedia data analysis and retrieval in order to provide effective tools for searching and discovering relevant content in real world applications. A multi-disciplinary approach will be used to bring together expertise from different areas in an effort to construct novel methodologies to fuse content and context information that would overcome current limitations. The MIESON project considers a user-centric approach; tools, methods and technology developed will focus on enhancing information retrieval and user experience in multimedia-enabled contexts.
Invito a presentare proposte
Vedi altri progetti per questo bando