Integrated Surveillance System for Real-Time Face Recognition and Image-Based Video Retrieval Using OpenCV and DeepFace

Trends of using intelligent surveillance systems have created a huge demand for real time face recognition technologies. Our work presents a logical sequence of ideas to put into practice the most effective methods including OpenCV, DeepFace, and Streamlit for developing a reliable, efficient, and secure surveillance system. The idea is to combine several possibilities: real time face recognition, usage of video stream in several threads, secure authorization of the user. A unique innovation in the system includes the ability to upload an image of a person to obtain video clips of the person in whatever capacity. This innovative capability mitigates problems related to existing surveillance systems, as it optimizes video search tasks; it is particularly valuable in forensic/ investigative use cases. With respect to the system's architecture, the approaches taken are fundamentally motivated by performance, low complexity, and scalability. Multithreading means that the system can process video streams from the several cameras in a parallel with faster and more stable rates. The novel video retrieval feature greatly minimizes the amount of time it takes to search for footage of a particular person, and this is done with acceptable latency. This makes this system ideal for use in large organizational structures that require real-time video stream processing because the structure of this system supports multiple streams at once. While adopting new technical solutions and the focus on the end-users the proposed solution provides a competitive outlook on actual scenarios of surveillance.