SDFC dataset: a large-scale benchmark dataset for hyperspectral image classification

Hyperspectral image (HSI) classification plays an important role in a wide range of remote sensing applications military and civilian fields. During past decades, significant efforts have been made on developing datasets introducing novel approaches to promote HSI classification, such that promising performance has achieved. However, existing generally pose following issues, including the limited categories annotated samples, lack sample diversity, as well low spatial resolution. These limitations severely restrict development evaluation data-driven models, especially deep neural network-based ones. In recent years, advances imaging spectroscopy provide us opportunity obtain hyperspectral data with high spectral resolution, therefore, this paper, we contribute large-scale benchmark dataset for conducting address issues raised by datasets, noted ShanDongFeiCheng (SDFC). The proposed SDFC is characterized (1) samples diverse categories; (2) resolution; (3) intra-class variance yet relatively inter-class variance, making task much more challenging it. We evaluated 10 classic traditional models SDFC, which results can be regarded useful baselines further experiments. Moreover, given state-of-the-art SpectralNet, selected it representation method, across analyze difference effects model induced different datasets. comprehensive review analysis representative both demonstrate advantages challenges our dataset, perspectives future studies.