what are the difference between bands and samples in indian pines hyperspectral dataset?

This question is closed and may reopen in the future if edited. for example: we have a very popular hyperspectral dataset named indian pines. consist of 220 bands total. in ground truth image its 145*145. i read some publications they often discuss about 10 samples per class, 20 samples per class etc. i am confused about it. please clear it with a simple way.

thanks

[John D'Errico]

This is NOT a question about MATLAB. Ask it on some site where it is on-topic.