[GRASS-user] classification losing class values of training data

[Samantha Goodchild-Brown]

Hi everyone

I have imported a shape file of Classification values, labels and RGB 
values into GRASS 6.4.1 in Linux ubuntu.
By setting the preferences key column first I manage to make a training 
raster map with correct classification values (double digit), but when I 
run the classification itself using i.gensig and i.maxlik, my 
classification results do not reflect the classification values but 
rather something like a feature count (single digit 1-4) which is 
something akin to the default cat values on import and not what I want.

I expected my classification to reflect the values of the training data 
and can find no way to force it.
What am I doing wrong?

Thanks
Sam