The growing need for better security and monitoring has exposed the shortcomings of traditional surveillance systems, which largely depend on manual observation and often fail to provide timely insights. This paper introduces a deep learning–based visual analytics framework aimed at improving the intelligence of modern surveillance systems through automated video analysis. The system uses computer vision and deep learning techniques to examine both live and recorded video streams, allowing real-time detection of objects, monitoring of activities, and recognition of unusual patterns. By converting raw video footage into useful visual information, the proposed approach minimizes the need for constant human supervision and enhances situational awareness. Experimental results show that the system performs reliably under different environmental conditions while offering better accuracy and faster response compared to conventional methods. Overall, this work presents a practical and scalable solution that supports the development of intelligent video analytics for security-oriented applications.