Large scale classification and retrieval

We are all aware of the information overload faced today, in social media, enterprise databases and public repositories. Digital cameras, digital imaging devices (e.g. in hospitals), digital video, professional document scanners, your Multi Function Device which prints / scans / copies in your home or office, all lead to digital image data. Classifying images or documents to assist workflow processes or manage media assets is a key technology. We research large-scale technologies that can deal with large amounts of media, and a high number of either generic or specific classes, without necessarily going to the cloud (cloud solutions are great, but not always appropriate due to cost or privacy restrictions for instance). On the other hand, instead of categorizing and grouping related content, image retrieval research focuses on retrieving specific content, based on image, text or combined queries. These issues are strongly connected with our research on image signatures, can we achieve these tasks, without requiring extensive computational power or adaptation?