This project was my final year university project.
An innovative smoke detection system prioritizing fast response times, affordability, and scalability. The system utilizes low-cost ESP32-CAM modules with built-in Wi-Fi and microcontrollers to capture live video streams, with basic image preprocessing handled on-device. These modules are strategically placed for comprehensive area coverage. It is not necessary to use an ESP32-CAM; any IP camera connected to the WLAN will suffice.
Smoke detection is performed using a lightweight YOLOv5s6 model processed on a central computer (with Jetson Nano as an option for real-world scenarios). The system delivers swift alerts via local alarms or smartphone notifications. Optional features for further development include GSM modules for uninterrupted transmission in weak Wi-Fi areas and solar power for extended outdoor deployments. Additional enhancements include flame location tracking, redundant detection frame removal, and intelligent delay mechanisms for improved accuracy and efficiency.