Active Learning-Driven Siamese Network for Hyperspectral Image Classification

Hyperspectral image (HSI) classification has recently been successfully explored by using deep learning (DL) methods. However, DL models rely heavily on a large number of labeled samples, which are laborious to obtain. Therefore, finding way efficiently embed in limited samples is hot topic the field HSI classification. In this paper, an active learning-based siamese network (ALSN) proposed solve problem First, we designed dual (DLSN), consists contrastive module and module. Secondly, view that difficult effectively sample under extremely labeling cost, adversarial uncertainty-based (AUAL) method query valuable promote DLSN learn more complete feature distribution fine-tuning. Finally, architecture, based inter-class uncertainty (ICUAL), construct lightweight pair training set, fully extracting information pairs improving accuracy. Experiments three generic datasets strongly demonstrated effectiveness ALSN for classification, with performance improvements over other related