Informedia: Lessons from a Terabyte+, Operational, Digital Video Database System
Chris Lichti, Christos Faloutsos, Howard D. Wactlar, Michael G. Christel, Alex G. Hauptmann · 1998
We describe the design decisions and lessons learned from Informedia, a digital video database system developed at Carnegie Mellon University and currently hosting over one terabyte of video data. Informedia is a research prototype for video retrieval by content, as well as a production system. The system has been implemented on a relational database, and it currently achieves less than five-second response times for ad hoc queries by video content on mid-range Pentium workstations. 1. Introduction The primary goal of the Informedia Project is the realization of full content search and retrieval from digital video, audio and text libraries. To achieve this goal, the project integrates speech, image and language understanding for library creation and exploration [1, 5, 11]. The Informedia News-on-Demand Project extends the research to the problems of continual yet timely library growth and focuses on news data [2, 7]. The Informedia Multilingual Project extends the research to the pro...