Demo paper: Savant - A Semantic Video Parsing and Tagging system

Hieu T. Nguyen, Hai Wei, Sameer Sheorey, Hoang Trinh

Abstract: We demonstrate a surveillance video tagging and search system that transforms input live raw video streams into semantic metadata structures through automatic object detection, tracking, classification and event recognition. The metadata generated through online video analysis and stored in spatial databases enable various types of user queries on the semantic content of archived video. They also allow for browsing and navigation through the video database effectively and efficiently. The system works with different application domains and camera setups.