Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Jan 13, 2011
Results of a methio for tracking individual targets in high density unstructured crowded scenes, a class of crowded scenes where the motion of the crowd at any given location is multi-modal over time. To this end we adopted the Correlated Topic Model (CTM) in which each scene is associated with a set of behavior proportions,where behaviors represent distributions over low-level motion features. Unlike some existing formulations, our model is capable of capturing both the correlation amongst different patterns of behavior as well as allowing for the multi-modal nature of unstructured crowded scenes. In order to test our approach we performed experiments on a range of unstructured crowd domains, from cluttered time-lapse microscopy videos of cell populations in vitro to videos of sporting events. In each of these domains we found that explicitly modeling the interrelationships between different behaviors in the scene allowed us to improve tracking predictions.