VID: A Comprehensive Dataset for Violence Detection in Various Contexts

To address the limitations of current datasets used for training automated crime and violence detection systems, we have created a new, balanced dataset consisting of 3,000 video clips. The dataset, which includes an equal number of violent and non-violent real-world scenarios recorded by non-professional actors, provides a more comprehensive and representative source for the development and assessment of these systems. Security and law enforcement professionals can use this comprehensive approach to analyze surveillance footage and identify pertinent incidents more efficiently.