NAYAN NIRIKSHAN – Violence Detection Using Surveillance Camera

NAYAN NIRIKSHAN - Violence Detection Using Surveillance Camera

Subhash B S, Sumanth Hegde, Nityanand Kumar, Harsha R, Bhagyashri Wakde

ABSTRACT
Move-net Pose Estimation is a method for real-time posture and movement detection of the human body using deep learning algorithms. Move-net Pose Estimation can be used to examine the video stream and find any unusual or violent behaviour when detecting violence using security cameras. A series of key points indicating the position of joints and limbs are produced by the system, and these key points can be used to recognize particular movements and postures that are suggestive of violence. This enhances the overall security of the monitored location by enabling the real-time detection and response to violent situations. Additionally, Move-net Pose Estimation can be tailored to recognize particular postures and motions in various settings and connected with other security systems like alarms and access control. But there are constraints.
Index Terms—

Move-Net Lightning, Yolo v3, Single-Pose Detection, Convolutional Neural Network, Violence and Non-violence activity (CNN).