Real-Time Fire & Smoke Detection from CCTV Footage: A Comprehensive Case Study on AI-Driven Visual Safety Systems

AThbissptaprear pcrte.sents a study of AI-driven, real-time fire and smoke detection using CCTV footage.It positions this approach as a visual-first safety solutionthat is proactive. The analysis shows a shift from traditional, reactive point-sensors to an “Internet of Eyes,” whichuses existing surveillance systems to provide critical situational awareness. The study looks at the evolution of detection algorithms, moving from earlier flawed models to modern deep learning object detectors like YOLO. Through three case studies—commercial, industrial, and smart city—this paper identifies practical deployment setups, such as Human-in-the-Loop and multi-sensor fusion. A significant contribution is the detailed discussion of the “false alarm problem” and its root cause in data scarcity, especiallyfor “incipient” fires. The paper also offers a technical analysis of key solutions, including Explainable AI, model optimization for edge computing, and the role of Generative AI. It concludes by summarizing the findings and suggesting a hybrid strategy for future development that emphasizes multi-sensor fusion, edge computing, and standardized regulatory frameworks.