VIRAT Video Dataset

event recognition from two broad categories of activities (single-object and two-objects) which involve both human and vehicles. (Sangmin Oh et al)