CA-MTransUNet Achieves 87% Accuracy in Cloud-Aware Forest Burned Area Mapping
• A new study in Big Earth Data introduces CA-MTransUNet, a Cloud-Aware Mixture-of-Experts Linear Transformer U-Net for forest burned area (FBA) detection using Sentinel-1 SAR and Sentinel-2 optical data. • The model attains the highest mean Intersection-over-Union (mIoU) of 87.00% with an inference speed of 6.26 ms, outperforming benchmarks despite cloud contamination. • This breakthrough addresses computational complexity in segmentation models, enhancing post-fire environmental monitoring critical for climate studies.
eurekalert.org
