Deep learning–based automated coastline extraction from satellite imagery: A Sentinel-2 benchmark dataset approach

The Sentinel-2 Coastline Training Dataset (SCTD) dataset and supporting code are publicly available, establishing a benchmark for coastline extraction in remote sensing and machine learning. This approach enables accurate monitoring of coastal erosion, assessment of engineering interventions, and support for climate adaptation planning. With multi-source and higher-resolution data integration, the framework can be extended to larger regions and more complex coastal types, representing a shift from instantaneous waterline detection to the extraction of geomorphologically meaningful “true coastlines”. For specific details, please refer to our literature materials "Deep learning - based automated coastline extraction from satellite imagery: A Sentinel-2 benchmark dataset approach".