Sentinel-1 Ground Truth Dataset for Coastline Extraction

Coastal environments are highly dynamic and vulnerable regions that require continuous monitoring to assess sea-level rise and support mitigation strategies for the protection of marine ecosystems and coastal infrastructure. Remote sensing, combined with deep learning techniques, offers a scalable, cost-effective, and timely solution for observing and analyzing these complex systems. This letter introduces a new dataset specifically designed for Synthetic Aperture Radar (SAR)-based coastline extraction, with the goal of fostering deep learning research in shoreline detection from SAR imagery. The dataset focuses on Southern European coasts and consists of freely available Sentinel-1 Single Look Complex (SLC) data paired with binary ground truth shoreline masks. To facilitate reproducibility and future expansion, the methodology used to construct the dataset from SAR and auxiliary data sources is provided, along with open access to the dataset itself. A comprehensive evaluation of the dataset’s applicability to deep learning-based shoreline detection is presented.