Identifying Anti-Social Activities in Surveillance Monitoring Applications using Deep-CNN based Algorithms

Abstract

Safety is the primary concern in present times. Crimes happen in public places and the criminal can quickly get away from the scene without anyone noticing him or any evidence against him. CCTV cameras are used for surveillance monitoring but they still need human supervision to operate and thus have a higher possibility of human error. So, in such cases, we need a machine to recognize such tasks and create evidence if it notices any such activity. Though many modern and advanced machine learning algorithms, processors, and CCTV cameras are available, but real-time detection is still difficult to achieve. Our work aims to create a system that identifies if any anti- social or abnormal activity is there or not from cluttered scenes. This works on Transfer Learning. We propose to use a Deep Convolutional Network (DCN), a state-of-the-art CNN model using the latest object detection technique YOLOv7. Using this in surveillance monitoring can be useful to reduce both the risk to human life and the rate of crime.