Convolutional neural network for smoke and fire semantic segmentation

In recent decades, global warming has contributed to an increase in the number and intensity of wildfires destroying millions hectares forest areas causing many casualties each year. Firemen must therefore have most effective means prevent any wildfire from breaking out fight blaze before being unable contain extinguish it. This article will present a new network architecture based on Convolutional Neural Network detect locate smoke fire. generates fire masks RGB image by segmentation. The purpose this work is help firemen assessing extent or monitor incipient real time with camera embedded vehicle. To train network, database corresponding images been created. Such allow compare performances different networks. A comparison best segmentation networks such as U-Net Yuan highlighted its efficiency terms location accuracy, reduction false positive classifications clouds haze. also efficient time.