Forest fires have a very bad impact on the natural environment and human beings. To protect the environment and enhance human safety, it is important to detect the source of a fire before it spreads. The existing fire detection algorithms have a weak generalization and do not fully consider the influence of fire target size on detection. To enhance the ability of fire detection of different sizes, ground fire data and Unmanned Aerial Vehicle (UAV) forest fire data are combined in this paper. To improve the detection accuracy of the model, a cosine annealing algorithm, label smoothing, and mult...
