Video Location Estimation System Receives Honorable Mention at MediaEval

^{October 9, 2012}
Speech Group researchers and their collaborators at UC Berkeley received distinctive mention at the MediaEval 2012 Workshop for their video location estimation system. Their system was named the most novel theoretical approach to addressing the workshop's placing task, which challenged participants to find the latitude and longitude of consumer-produced videos and photos pulled from Flickr and YouTube. The ICSI/Berkeley system is the first multimodal video location estimation system that uses all three modalities available in videos: visual, acoustic, and textual information, such as user-annotated descriptions. The annual MediaEval Workshop is dedicated to evaluating new algorithms for multimedia retrieval, which allows users to search for media using text, audio, and visual cues. The ICSI/Berkeley system was developed by Jaeyoung Choi and Gerald Friedland of the Speech Group and graduate student Venkatesan Ekambaram and Professor Kannan Ramchandran of UC Berkeley. Read the award-winning paper.