Avoiding Overfitting When Applying Spectral-Spatial Deep Learning Methods on Hyperspectral Images with Limited Labels | IEEE Conference Publication | IEEE Xplore