Due to the abundance of particular material in HSI image. Data corresponding to each material(class label) is biased.
Some papers have written take fix percentage(15 %) of labelled data for each class.But still the data will be biased
How to use HSI data for training such that there is no bias for better classification result?