Smart Fire-Watch Monitoring System
A collaborative drone-ready wildfire monitoring platform for the Karabük region that combines weather-based FWI risk prediction, real-time fire/smoke detection, alert management, and operator dashboard views.
Overview
Overview
As part of a team graduation project, we designed and built an automated, drone-ready wildfire prediction and detection platform tailored for regional forest monitoring in Karabük. Team Architecture & Technical Contributions: Two-Stage Machine Learning Pipeline: Collaborated on engineering a prediction pipeline using 34 custom weather features to calculate Fine Fuel Moisture Code (FFMC) and Fire Weather Index (FWI) values, achieving 92.9% high-risk classification accuracy and 90% recall. Real-Time Computer Vision: Integrated custom YOLOv8 models for visual smoke and fire detection across streams from field cameras and aerial drones. Backend & Integration: Contributed to a scalable FastAPI backend supporting 30+ endpoints, automated weather data ingestion from Open-Meteo API, manual/scheduled risk evaluation runs, and integration with an interactive Next.js/TypeScript operator dashboard for live alert dispatching.
Project Gallery