Deep Learning for Remote Sensing to Improve Flood Inundation Mapping
Researchers have developed a new cloud-removal framework for flood imagery using Denoising Diffusion Probabilistic Models and a Masked Diffusion Transformer architecture. This method aims to improve flood inundation mapping by generating cloud-free satellite images, which are crucial for disaster risk management. The model reconstructs obscured regions by leveraging self-attention mechanisms and masked token modeling, preserving hydrological consistency and spectral signatures for accurate water detection. AI
IMPACT Enables more reliable, continuous satellite observations for disaster risk management and flood-related decision making.