Instant analysis; instant results
ObjectVideo is researching solutions that help simplify the analysis and subsequent reasoning of media. The ultimate goal of our research efforts is to automatically determine media content, i.e. express media content in terms of 'who,' 'what,' 'when,' 'where,' 'why' and 'how.' Currently this task requires sensitive evaluation of contextual clues and mundane searches through vast data repositories. ObjectVideo is currently developing solutions that help with tedious tasks such as data upload, search, and reporting, and with challenging tasks such as content categorization (triage), deep analysis and response support.
Pixels become data
ObjectVideo's search capabilities help to bridge the gap between pixels and textual summaries of video content, enabling the user to search for complex and context-sensitive events. The system utilizes scene descriptors, object and event meta-data that are generated by a video analysis engine. The recognized semantic contexts generated from automated scene analysis algorithms are converted to a plain text description that are indexed and stored in a COTS database. The query and retrieval system accepts keyword, natural text, visual concept or a video clip as a query and provides the most relevant image, video and text description from the database.