FDPS: A YOLO-Based Framework for Fire Detection and Prevention

Accidental fires or explosions pose a significantly threat to human life and social safety, making them a major concern for humanity. In this study, we propose FDPS, an inexpensive and efficient machine learning-based framework for the early detection and extinguishing of small fires using a minimal amount of Internet of Things equipment, such as a NodeMCU, camera, and pump. The system functions through two primary phases: fire detection and fire extinguishing. During the initial phase, the camera rotates horizontally while continually analyzing images that were collected and processed by a pre-trained YOLO model to detect the presence of fire. The second phase activates when the fire is detected, and the system transmits signals to adjust the camera’s angle and activates the water pump, which then extinguishes the fire. Our experiment results demonstrate that the proposed framework can accurately detect fires, achieving an average precision of nearly 90% and running inference in about five milliseconds. By detecting fires quickly, emergency responders can be alerted to respond promptly, potentially reducing property damage and saving lives.