An Effective Real-Time Surveillance System for Fire and Smoke Detection Using CNN

Fire disasters are the most dangerous and lethal events that can cause social, economic, life losses. Early detection of fire or smoke is crucial to facilitate intervention in time to avoid large-scale damage. Hence, the effective utilization of embedded devices enhances the performance of the overall surveillance system. We proposed an efficient, low-cost, real-time, memory-optimized method for surveillance systems using the YOLOv4-tiny model for early fire and smoke detection. The proposed method provides the implementation of fire and smoke detection systems for real-world applications where the model could run on low-cost hardware like 1.44 GHz processor devices. The experimental results show that the developed surveillance system can detect fire and smoke in real-time.