CloudCompare forum

Hello,

I've trained a classifier and I now want to classify (large) point clouds.
Before classifying, I'm computing the features with python using the following command : cc.q3dmasc(clouds, PARAMETERS_PATH + parameter_file, only_features=True, verbose=True)

It seems like the command can't compute the features. It doesn't show any error but it saves files with the following extensions (instead of .sbf):

Do you have any recommendations on how to classify a large point clouds ?
Should I try to reduce as much as possible the number of parameters ?
Do I have to maybe subsampled the point cloud, classify the samples and then merge them back together ?

Thanks

Do you have the ability to test with the GUI version of the plugin first? It's unclear if it's related to the Python interface, or to the 3Dmasc plugin itself.
Especially since I'm pretty sure the authors use it on very large clouds.

Cloud you provide your parameter file and your data, or at least a sample of it? I never saw such an error.

If you have strategies to try, I would say that you should try them to gain some experience. In terms of classification, practice is a good thing. Begin with a small subset of your data before letting the software work on a larger part of it.

Hello,

I can't test it with the GUI because I'm using scikit to train my classifier (if I've understand right the plugin works with the OpenCV models).
I'm going to try switching to OpenCV as well to fix my issue.

I've tried using a smaller point cloud and using subset of data and it worked fine. I'll send you the parameter file and data in private.

I'll try other strategies and let you know !

Thanks for your help

So the command can compute the features. Does it close this post?

It does work on small samples but it still isn't working on bigger samples of data.
I'm now splitting and then merging the clouds. I found a way to deal with it but not to properly fix it.
I'm not planning on fixing this issue so I guess we can close this post.