A multi-task learning semantic segmentation approach is employed for targeting both wildfire delineation and burn severity estimation. By exploiting a large dataset of images from past wildfires and integrating both tasks into a single network can exceed the state-of-the-art results on this topic.