Improving Flood Insights: Diffusion-based SAR to EO Image Translation (Papers Track)

Minseok Seo (si-analytics); YoungTack Oh (SI Analytics); Doyi Kim (SI Analytics); Dongmin Kang (SIA); Yeji Choi (SI Analytics)

Paper PDF Poster File Recorded Talk NeurIPS 2023 Poster Cite
Disaster Management and Relief Earth Observation & Monitoring


Driven by the climate crisis, the frequency and intensity of flood events are on the rise. Electro-optical (EO) satellite imagery is commonly used for rapid disaster response. However, its utility in flood situations is limited by cloud cover and during nighttime. An alternative method for flood detection involves using Synthetic Aperture Radar (SAR) data. Despite SAR's advantages over EO in these situations, it has a significant drawback: human analysts often struggle to interpret SAR data. This paper proposes a novel framework, Diffusion-based SAR-to-EO Image Translation (DSE). The DSE framework converts SAR images into EO-like imagery, thereby enhancing their interpretability for human analysis. Experimental results on the Sen1Floods11 and SEN12-FLOOD datasets confirm that the DSE framework provides enhanced visual information and improves performance in all flood segmentation tests.

Recorded Talk (direct link)