End-to-end Multiplayer Violence Detection based on Deep 3D CNN

Numerous behavior recognition researches have focused on UCF-101 video dataset, such as sports, cooking and other simple routines. Yet these studies are less useful in real-life surveillance scenarios. Violence detection in crowded scenes (such as shopping malls, banks, and stadiums) is significantly important but little research has been done. Based on this situation, this paper proposes a multiplayer violence detection method based on deep three-dimensional convolutional neural network (3D CNN), which extracts the spatiotemporal feature information of multiplayer violence. Our method directly detects violence in an input video by end-to-end. The experimental results show that the accuracy of our method is higher than the methods of artificially extracting features in violence detection.