Abstract: Data quality is essential for training deep learning models, and recently, the challenge of noisy labels in hyperspectral image (HSI) classification has attracted considerable attention.