IoT and ML-Driven Smart Fire Alarm and Crowd Tracking

In response to the escalating global fire incidents, particularly in densely populated areas, this research pioneers a vision-based fire detection system to overcome the limitations of conventional methods. Integrating machine learning and deep neural networks within a Raspberry Pi framework, the study employs MobileNet SSD for precise people counting and ResNet101 for robust fire detection, fine-tuned via transfer learning with diverse datasets. The system’s core, integrated with a Kinect camera, enables swift transitions between fire and human detection. Rigorous real-world testing validates its reliability, emphasizing its efficacy in fire detection and crowd tracking. Anticipated outcomes emphasize early fire detection, adaptive resource allocation, and cost-effective applicability in diverse settings. This innovative approach marks a significant stride in fire safety, leveraging IoT and machine learning for proactive emergency responses and laying the groundwork for continued advancements in safety protocols, promising a transformative impact in mitigating fire-related risks and enhancing urban safety measures.