S3CD: SAR-Optical Sea-Land-Shore Scene Classification Dataset

The study areas encompass the coastal region at the Hebei-Tianjin border (characterized by flat mudflats and intensive human activity) and the Fujian coastal area (dominated by rocky shores with complex geomorphology). The two regions differ significantly in landform type, coastline morphology, and intensity of human activity, supporting evaluation of model generalization and robustness.The dataset uses Sentinel-1 IW GRD data and Sentinel-2 Level-1C data acquired in October 2024.

Sentinel-1 data includes VV and VH polarizations at 10-meter spatial resolution, preprocessed through orbit correction, thermal noise removal, radiometric calibration, terrain correction, and decibel conversion. Sentinel-2 data uses bands B4 (red), B3 (green), and B2 (blue) to compose true-color images. Image pairs were selected with an acquisition time difference within 3 days, geometrically co-registered, and uniformly cropped into 256×256 pixel patches to ensure spatial consistency.The dataset contains 8,649 precisely labeled images across three categories: Sea, Land, and Coastal.