TrecVid Competition

Previous

Students

Joriz DeGuzman, Computer Science

Jim Kleban, ECE
Emily Moxley, ECE
Swapna Joshi, ECE
Stephen Mangiat, ECE
Jie-Jun Xu, Computer Science

Faculty Advisors

B.S. Manjunath, Elec & Comp Engineering

Next

Abstract

The UCSB TRECVID team is leveraging the multimodal nature of video to research the area of video search and retrieval. Specifically, the team is focused on detecting the presence of certain concepts in a video shot, timely and effective retrieval of video pertinent to a multimodal query, and efficient video summarization. The team is loosely split into three overlapping areas of expertise--text, image, and audio features--that are fundamental to achieving these goals in video. Characteristics derived from these features are fused and presented using machine learning techniques and a search engine interface on the web. The team participates in the annual international TRECVID benchmarking collaboration sponsored by NIST sponsored that evaluates performance on video search-related tasks.