Real-time Violence Detection in Video Streams
Surinder Kaur, Ajay Dureja, Mahesh Kumar, Mohit Dayal, Shyla · 2024
Although rarely utilized to prevent crime or respond to it in real-time, CCTV is a significant piece of evidence in court. Despite having accuracy rates that are close to 85%, present approaches lack precision and memory requirements. For the majority of security systems, automatic violence detection is still an unresolved problem. Due to the increase in violent crime, thorough surveillance systems are now required to spot potentially dangerous situations and deal with violent encounters. Modern methods for detecting violence in surveillance recordings include the extraction of auditory features, spatiotemporal analysis, or the construction of optical flow motion vectors. The feature extraction techniques and object detection techniques are implemented in a dataset for detection of violent activities in the proposed dataset.