CANUPO differences between GUI/Command Line

[cmackSIO]

Hello all,
I'm attempting to classify vegetation on a LiDAR survey of some cliffs using CANUPO - I've trained a classifier that works well but I get much different results when classifying in the GUI vs through command line. Using the GUI, I set my confidence level to 0.8 and I get this:

Screenshot 2022-12-15 153352.png (256.01 KiB) Viewed 2065 times

This is the result I'm looking for, however, when I use the command line mode with the same classifier and the same .8 confidence threshold, I get this:

Screenshot 2022-12-15 153155.png (707.29 KiB) Viewed 2065 times

Which is much different and clearly is overclassifying vegetation by a lot. I'm wondering if anyone else has run in to this issue and found a solution. My command line call is:

Start-Process -Wait "C:\Program Files\CloudCompare\CloudCompare.exe" -ArgumentList '-silent', '-clear','-auto_save', 'off', '-c_export_fmt',
'asc','-o','-global_shift', $shift, $c1,'-canupo_classify','-use_confidence', '0.8', $canupo,'-save_clouds','file',$output_string
(where $canupo contains the path to my .prm file)

I'm wondering if maybe the arguments are being passed in the wrong order so my confidence threshold is not being recognized? Though the example in the docs is:
'CloudCompare -O cloud1 ... -O cloudN -CANUPO_CLASSIFY (-USE_CONFIDENCE 0.9) classifier.prm'
so that would seem to be the right order. I'm not using any scalar field in either classification though I could add intensity to the command line call and see if that helps - regardless I'm not sure why the behavior of the two would be so different. Any thoughts would be appreciated, thanks!

[daniel]

If the parameter is properly read, you should see in the (command line) console:

Code: Select all

Confidence threshold set to XX

I wonder if another issue might be with the Global Shift? Can you try with the 'AUTO' option?

[cmackSIO]

I do see the console registering the confidence level so that's not the problem. I'm now able to get much more similar results but they're still slightly different. In command line mode I get this:

Screenshot 2022-12-19 140906.png (200 KiB) Viewed 1988 times

While in the GUI I get this:

Screenshot 2022-12-19 141108.png (210.21 KiB) Viewed 1988 times

Both are .8 confidence with the same classifier - the command line results are good enough now that I can use them but I just want to understand why they're different - is there anything else I might be missing? Thanks for your help

[daniel]

Hard to tell... What about the Global Shift, is it the same? Or maybe it's just a display thing?