Wo: Freie Universität Berlin, Seminarraum SR049, Takusstr. 9, 14195 Berlin (location)
Wann: 14:00 - 15:30
Wer: Dr. Gerald Friedland, Direktor Audio and Multimedia Group, am International Computer Science Institute der University of California, Berkeley
Drinking from a Multimedia Firehose: Multimodal Large-Scale Analysis of Consumer-Produced Videos
Consumer-produced videos are the fastest-growing type of content on the Internet. YouTube alone claims that 100 hours of video are uploaded to its web site every minute. Because these videos capture parts of the world, they are potentially useful for global-scale empirical research. A major prerequisite to making social media videos usable for "field studies" is efficient and unbiased (e.g., keyword-independent) retrieval. More importantly, retrieval needs to go beyond simply finding objects to detecting more abstract concepts, such as “baby learning to catch a ball” or “animal dancing to music”. Research on such a diverse corpus requires the creation of methods that exploit as many cues as possible from different modalities. ICSI has begun using novel acoustic methods to complement computer vision approaches evaluated on the recently released Yahoo-Livermore-ICSI corpus (1M videos). This talk summarizes ICSI’s research and progress in this area.
Gerald Friedland is the director of the Audio and Multimedia group at the International Computer Science Institute, a private, non-profit research lab affiliated with the University of California, Berkeley. He is also lecturing at the EECS department. His research interests focus on multimedia content analysis, its applications, and social implications. Current research projects include large-scale automatic geo-location of videos, concept-based video retrieval, and the privacy implications of social-media search. Dr. Friedland has published more than 200 peer-reviewed articles in conferences, journals, and books. His co-authored text book on Multimedia Computing appeared in August with Cambridge University Press and his co-authored book on "Multimodal Location Estimation of Images and Videos" will appear with Springer in October this year. He is the recipient of several research and industry recognitions, among them the European Academic Software Award and the Multimedia Entrepreneur Award by the German Federal Department of Economics. He also lead the team that won the ACM Multimedia Grand Challenge in 2009. Dr. Friedland received his doctorate and master’s degree in computer science from Freie Universitaet Berlin, supervised by Prof. Raul Rojas, in 2002 and 2006, respectively.