During the ISD Demo and Exhibit Teasers, 10:30-12:00 on Friday September 2nd, Jacob Becker will be presenting ‘A large-scale Benchmark Dataset for Event Recognition in Surveillance Video.’
Abstract
We introduce to the surveillance community the VIRAT Video Dataset, which is a new large-scale surveillance video dataset designed to assess the performance of event recognition algorithms in realistic scenes. Exemplary video clips along with annotations and diverse event statistics will be showcased during the demo. The dataset includes videos collected from both stationary ground cameras and moving aerial vehicles. We expect the dataset to further research in continuous visual event recognition(CVER), where the goal is to both recognize an event and to localize the corresponding space-time volume from large continuous video. This is far more closely aligned with real-world video surveillance analytics needs than the current research which aims to classify a pre-clipped video segment of a single event. Accurate CVER would have immediate and far reaching impact in domains including surveillance, video-guided human behavior research, assistive technology, and video archive analysis.
Physical Event