-r#(0,1) is not working to produce unbalanced classes.

[GoogleCodeExporter]

What steps will reproduce the problem?
1.Command line:

 /home/eckleyd/CleanPlace/iicbu/wndchrm/branches/wndchrm-1.30/wndchrm test -r#0.9 -j10 -n20 -l /home/eckleyd/ICAWormImaging/Working/NewNotRotated/Newer.fit /Images/dme/z_happy_copy/NewerUnbalanced.html

What is the expected output? What do you see instead?
expected about 450 training images for Background, 200 for Body, etc.
got balanced classes, see below.

Newer
994 Images.
Class   Value   Images
BackgroundNR    0   526
BodyNR  0   223
HeadNR  0   135
TailNR  0   110

Images for training and testing (per-split)     BackgroundNR    BodyNR  HeadNR 
    TailNR  total
Testing     10  10  10  10  40
Training    99  99  99  99  396

What version of the product are you using? On what operating system?
wndchrm 1.30.182, linux

Please provide any additional information below.
I'd like to use my hard-earned images for training a classifier.
I would also like to test the idea that unbalanced classes have a detrimental 
effect on my 4 class classifier. 
As it stands this test is impossible and I cannot use more than the number of 
images in my limiting class.

Original issue reported on code.google.com by dmarkeck...@gmail.com on 1 Apr 2011 at 3:19

[GoogleCodeExporter]

This issue was closed by revision r212.

Original comment by i...@cathilya.org on 20 Apr 2011 at 8:31

• Changed state: Fixed

[GoogleCodeExporter]

Fixed.  The split_ratio wasn't being propagated properly.  Also, if unbalanced 
training is specified, then testing is done using remaining images in each 
dataset (unbalanced).  Balanced testing can be specified with unbalanced 
training using the -j switch.

Original comment by i...@cathilya.org on 20 Apr 2011 at 8:59