Abstract: In modern times, security and protection have become paramount concerns. This research paper introduces an innovative approach to fortify security in premises through a sophisticated smart surveillance system. Supported by a Django backend and empowered by various trained models, the system enables real-time monitoring of entries and exits, automated counting, and anomaly detection. It aims to revolutionize security and safety across diverse environments by automating tracking of individuals and objects, monitoring for suspicious activities, and early detection of fire hazards. The paper discusses the current state of technology in smart surveillance systems, covering advancements, challenges, and demerits. It outlines the objectives, motivation, and challenges of the project, emphasizing the importance of algorithm selection and fine-tuning, real-time processing, accuracy and reliability, adaptability to environmental changes, and addressing privacy concerns. The abstract underscores the project’s commitment to delivering a comprehensive solution that leverages technology to enhance accuracy, efficiency, and security in premises.

Keywords: Computer Vision, Image Processing, YOLOv3-Tiny, Real-time object detection, Convolutional Neural Networks (CNNs), Non-Maximum Suppression (NMS), Structural Similarity Index (SSIM), Haar Cascade Classifier, classifier, Anomaly Detection, Object Recognition, Emergency Response, Fire Detection, Object Monitoring, Surveillance Technology.

| DOI: 10.17148/IARJSET.2024.11646