BdWaterSurfaceSet : Automated Land Watersurface Detection in Bangladesh

BdWaterSurfaceSet is a satellite imagery dataset designed for automated detection of inland water surfaces in Bangladesh, supporting research in remote sensing, environmental monitoring, and computer vision. The dataset contains image samples representing water bodies such as lakes, canals, ponds, and rivers, as well as non-water land regions collected from diverse geographic locations and surface conditions. It was developed through a structured preprocessing pipeline that extracts region-level samples from raw satellite images while maintaining balanced and reliable class representation. The dataset is divided into training, validation, and test subsets to ensure fair benchmarking and reproducible evaluation. In addition to classification labels, the dataset also provides annotations in COCO Segmentation Format, enabling its use for image segmentation, object detection, and instance segmentation tasks. BdWaterSurfaceSet can therefore support binary classification, segmentation, land-cover analysis, and change detection studies, providing a valuable benchmark resource for developing deep learning models for water surface recognition, particularly in South Asian environments where such public datasets are limited.