Violence Detection in Real Time for Surveillance

In an interconnected world, the need for advanced surveillance systems capable of detecting criminal offenses and violence is necessary. This paper aims to create a real-time surveillance system using the latest available technology, which focuses on the detection of criminal actions like human fights from surveillance camera footage. The paper makes use of YOLO (You Only Look Once) model, an advanced ML method, which is known for its speed and accuracy in object and action detection. The system employs alert mechanisms to respond to suspicious activities. It extracts its frames and distinguishes between violence and non-violence using data labelling. The YOLO model is fine- tuned to identify the patterns and characteristics of violent actions. Once the model is integrated with the surveillance system it can raise alerts about detected offenses. Human operators can review alerts and take appropriate actions. As YOLO enables the system to provide real-time alerts, It could analyse the videos live and inform concerned security or authority. This helps in taking immediate action on incidents. Several datasets are publicly available or pre-trained models for object detection and action crime recognition can be used. Thus, it supports a diverse range of data and performs various real time activities.