CloudCompare forum

Hi, I'm trying to combine some scalar fields of points and the multi-scale dimensionality metrics from CANUPO together for classification or unsupervised clustering of points. Is there anyway I can export the multi-scale metrics from qCANUPO in CloudCompare so that I can use them along with the other scalar fields in a preferred clustering algorithm? I tried the original CANUPO program from http://nicolas.brodu.net/en/recherche/canupo/ to export it. But I found that the speed of that program is somehow much slower than the qCANUPO implementation in the CloudCompare. Not sure if I did something wrong with the original program. I did use the same amount of core points for metric calculation.

If possible, can I borrow the source code of the plugin qCANUPO and add export function or wrap the qCANUPO implementation into a standardalone but faster CANUPO. I checked the repo of CloucCompare on github but didn't find the source code of qCANUPO.

Have you got any thought on this? Thanks for any comment or suggestion.

Zhan.

Indeed the qCanupo plugin is multi-threaded while the original program is not.

And sadly the code is not public yet. The code of the original program is on the contrary.

Moreover pay attention that the dimensionality is not a single value but in fact 2 values for each scale.

Daniel, thanks for your response. Sorry that the code is not public yet but I got the confirmation that the qCANUPO is indeed faster than the original one for a reason! I realize that the dimensionality is 2 values for each scale. I currently wrap the original C++ program in python and use these exported functions to read msc files. It works fine except coming to large files which are what I really need to work on. I'll see if I can find any other compromise between speed and data size.

Thanks,

Zhan.

Daniel, but may I suggest that the qCANUPO add a function that the classification can take in not only the dimensionalty but also some scalar field of points? I think sometimes this will be very helpful when some attributes of points, such as intensity that can boost the classification accuracy.

Zhan.

I believe it's already in the pipe (as well as colors, etc.). However you should tell that to Dimitri Lague, the author of the plugin ;)