High-resolution UAV RGB imagery and segmentation masks for biological soil crust mapping in a mining-impacted dryland ecosystem

This dataset contains high-resolution UAV RGB image patches and paired binary segmentation masks for biological soil crust mapping in a mining-impacted dryland ecosystem in the Mu Us Sandy Land, northern China. The dataset includes 1,640 RGB image patches, 1,640 paired binary masks, predefined training/validation/test split files, and metadata files describing image-mask pairing, class distribution, acquisition parameters, and integrity checks. The masks use a binary label definition: 0 = background and 1 = biological soil crust. The predefined split contains 1,312 training patches, 164 validation patches, and 164 test patches. The dataset is intended for semantic segmentation benchmarking, UAV-based ecological monitoring, and mine restoration assessment.