 The proposed method is a semi-supervised approach to classifying hyperspectral images. It uses a three-dimensional bilateral filter to extract spectral spatial features from the images, then trains a generative adversarial network, GAN, on these features to generate additional data points for the classifier. This allows the classifier to be trained with fewer labeled samples while still achieving good performance. This article was authored by Jihee, Pan Lu, Yi Wen Wang, and others.