In the domain of hyperspectral image (HSI) classification, the majority of deep learning methods have necessitated a substantial number of manually annotated samples to achieve outstanding results.