Engineering Smart Detection Systems: Leveraging Neural Networks for Crime Prevention

In recent years, numerous techniques for surveillance security have emerged to enhance public safety and mitigate the risks associated with deviant human activities. This research addresses deviant activity detection by proposing an advanced surveillance system capable of identifying actions such as smoking, harassment, and fighting. The system employs a hierarchical technique and utilizes Convolutional Neural Networks (CNN) to analyze captured images and detect deviant behaviors. A diverse dataset was compiled and augmented with techniques like rotation, zoom, and horizontal flip to improve model robustness and performance. The proposed model achieved an accuracy of 93.33% in identifying deviant classes during testing and validation. The system significantly reduces the need for human intervention by automating the detection process, thereby enhancing response times and overall security. Extensive empirical observations validate the effectiveness of the system in various environments and conditions.