att_abstract={{This paper presents a general surveillance event detection system that participated in the interactive Surveillance Event Detection (iSED) task of TRECVID 2013. In the proposed system, a set of spatio-temporal features including Space-Time Interest Points (STIP) and Dense Trajectories are extracted, and a sliding temporal window is employed as the detection unit. Fisher Vector is used to encode low-level features as the representation of each sliding window. Both feature-level and decision-level fusions are used to combine multiple features. In order to deal with the highly imbalanced nature of surveillance data, the system performs detections using the CascadeSVMs algorithm according to each specific event and camera view. Two different interactive environments are evaluated, one focuses on high throughput and the other includes related result expansion. In the primary run evaluations, our system ranks the top in 2 out of 7 event detection tasks. }},
	att_authors={zl3194, ez2685, dg1597, bs1261},
	att_copyright_notice={{The definitive version was published in  2013. {{, 2014-03-01}}{{, http://www-nlpir.nist.gov/projects/tvpubs/tv.pubs.org.html}}
This notebook paper is for circulating at the TRECVID workshop only. The final version will be made public in 3/1/2014.}},
	att_tags={Surveillance Event Detection, CascadeSVM, Space-Time Interest Point, Dense Trajectories, Fisher Vector},
	author={Zhu Liu and Eric Zavesky and David Gibbon and Behzad Shahraray and Xiaodong Yang and YingLi Tian},
	institution={{TRECVID workshop by NIST}},
	title={{AT&T Research at TRECVID 2013: Surveillance Event Detection}},