HiEve, short for Human-in-Events, is designed to understand human motion, gestures, and actions in various real-world events, especially crowds and complex events, including getting on and off a subway, collisions, fights, and earthquake escapes. HiEve is a challenging dataset for human-centric video analysis.

The dataset includes the current largest number of poses (> 1M), the largest number of complex event action labels (> 56k), and the largest number of long-term continuous trajectories (average trajectory length > 480).