Multi-Modal Attention for Automated Disaster Damage Assessment Using Remote Sensing Imagery and Deep Learning
Researchers have developed a new deep learning framework to automate disaster damage assessment using remote sensing imagery. The system fuses pre- and post-disaster satellite data with a multi-modal attention mechanism to classify buildings into four damage levels, achieving 94.90% accuracy. This approach significantly enhances the speed and precision of damage evaluation, aiding emergency response efforts. AI
IMPACT Automates disaster damage assessment, improving emergency response speed and accuracy.