OBJECT DETECTION BASED SECURITY SURVEILLANCE SYSTEM

Abstract

OBJECTIVE: This project aims to enhance security and surveillance systems by integrating multiple detection models, including object detection, face detection, face recognition, and motion detection, into a unified framework. The primary objectives of the project are to identify potential threats, detect anomalous behavior, and provide access control through automated processes. METHODS: To achieve these objectives, they employ state-of-the-art models such as YOLO- NAS for object detection and face detection. Additionally, motion detection is implemented using background subtraction methods. These models run simultaneously on edge devices, reducing latency and enabling real-time processing. The metadata generated by each model is sent to the cloud for further analysis and automated tasks, such as notifications and identity verification. FINDINGS: After training the model, it achieves the result of an mAP of 50.75%, f1 score of 5.51%, precision of 2.88% and recall of 82.7% in 100 epochs and inference on a video is of around 10-12 fps. NOVELTY: The novelty of this project lies in its comprehensive approach to security enhancement, leveraging cutting-edge detection technologies and edge computing to create a smarter and more efficient surveillance system.