DESIGN AND DEVELOPMENT OF AN AGENTIC AI FRAMEWORK FOR REAL-TIME ANOMALY DETECTION USING OPEN-SOURCE MODELS

This thesis focuses on the design and development of an Agentic AI-based framework for real-time anomaly detection through video stream analysis. The work involves the integration of open-source Deep Learning models for anomaly detection and the implementation of strategies to reduce false alarms. The system performance will also be evaluated in terms of accuracy, robustness, and computational efficiency.