An Intelligent System to Detect Violent Mob Activities

The Paper here discusses and further provides users with a solution to tackle the unending problem that is Mob Violence. The approach involves Human Pose estimation technique to detect human skeleton poses and Action recognition using DNN to detect violent actions, especially involving a mob. The introduced approach will aid the field of mob violence surveillance and violence detection systems by accurate pose estimation and easily distinguishable actions. Violent Behaviors are hard to trace given their nature and rapid changes in actions. Hence, taking into consideration what a violent act is, the dataset is classified into 9 different sub-actions namely 'walk, wave, punch, kick, stand, run, jump, squat, and sit'. This is because every violent action can be broken down into pieces of information that can be easily processed. The Results provided us with sufficient accuracy so as to implement the system.