A National-scale Sandy Beach Dataset for India Derived from High-resolution Satellite Imagery and Deep Learning

This dataset provides the georeferenced delineation of sandy beach extents along the Indian coastline, derived from IRS ResourceSat-2/2A LISS-IV multispectral imagery using a U-Net deep learning model. The dataset captures fine-scale sandy shoreline features and serves as a foundational resource for coastal geomorphology, shoreline monitoring, and sustainable coastal zone management. The sandy beach polygons were extracted through a deep learning workflow implemented in PyTorch, employing a U-Net segmentation model trained on manually annotated coastal sites representing diverse geomorphic and sedimentary settings. The input imagery comprises 5.8 m spatial resolution LISS-IV data with green, red, and near-infrared bands, supplemented by derived indices such as NDVI, Green–NIR ratio, and composite intensity to improve feature separability. The shapefile provides polygon representations of mapped sandy beaches across the Indian coastline and is projected in WGS 84 geographic coordinates (EPSG:4326). Users can integrate this dataset within GIS environments for visualization, spatial analysis, and model validation. Temporal coverage of the imagery used for mapping spans 2021–2024. Users should note that very small or seasonally transient sandy patches may not be captured due to the 5.8 m spatial resolution of the source data. This dataset supports applications in coastal mapping, blue economy planning, sediment dynamics research, and shoreline change assessment.