Interpretation of human behavior from video, is essential for human-machine interaction, public safety and more. Together with the national police and the city of Arnhem, we researched the feasibility of aggression detection during nightlife.

We detected >40 incidents during a test period of 3 months, with one false positive per hour. The algorithm is based on machine learning and motion analysis. It looks for wild motion patterns between people (interactions).
The model was trained with few examples, which is key for learning to spot rare events.