An Abridged Investigation of Deep-Learning-based Video Crime Detection Systems

Crime detection in surveillance video represents an essential process to minimize crime activity before it happens. Furthermore, video crime detection systems are essential to prevent harm to private/public property, save the lives of victims, evade all-time strain, and develop a peaceful community. Additionally, these systems can be beneficial in predicting potential terrorist activities. Recently, video crime classification and detection systems dependent on deep learning approaches become an attention-grabbing area of research. This paper has exhibited an abridged investigation of recently prevailing deep learning-based systems for detecting crime activities in surveillance video to automate hazardous events and enable law enforcement organizations to provide efficient steps toward public safety. A comparative analysis, commonly utilized benchmark crime datasets, and evaluation metrics are presented in this investigation. Moreover, an implementation of effective deep learning approaches is also presented to specify the best systems for the specified application.