In recent years, wildfire has become a significant issue in California, causing enormous damage to the residents and the environment. To prevent the widespread of wildfires, applying machine learning algorithms to detect and forecast wildfires is a feasible path. This project will acquire video captures from surveillance cameras, detect the source of smoke and fire using machine learning algorithms, and send back the result to a wildfire detection network for further reactions and analysis.
The dataset is collected by the High-Performance Wireless Research & Education Network (HPWREN) of UCSD and labeled by the SDSC researchers. HPWREN cameras are installed in remote locations to surveillance areas that are not well-covered by other technologies. The dataset contains screenshots from video snippets of wildfire events. The images are labeled positive or negative, depending on whether the screenshots contain fire or smoke.