Abstract: In hyperspectral cross-scene classification tasks, it is often challenging to obtain target domain samples during the training phase. Therefore, models need to be trained on one or multiple ...