ABSTRACT

The project aims to develop an advanced surveillance system using video data to classify and detect different activities in real-time. The system classifies videos into three primary categories: Normal, Violence, and Weaponized. To achieve this, a hybrid deep learning approach is utilized, combining Convolutional Neural Networks (CNN) for spatial feature extraction and Gated Recurrent Units (GRU) for temporal sequence modelling. The CNN efficiently extracts spatial features from individual frames, while the GRU handles the sequential nature of video data, identifying patterns over time. The system is trained on a diverse video dataset containing various real-world scenarios, ensuring robustness and accuracy. The proposed solution aims to provide real-time monitoring and alerting for security purposes, improving public safety and preventing potential threats. By leveraging the power of deep learning, the model offers high classification accuracy and efficiency, making it a promising tool for modern surveillance applications. This project also explores optimization techniques to enhance performance, focusing on the practical deployment of the system in surveillance environments.

Keywords: Surveillance System, Video Classification, Deep Learning, Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), Activity Detection, Real-time Monitoring, Threat Detection