Sorry, we didn't log in recently. We are pleased to see qCSF is useful to your work.
To your question, could you give more details, such as figures or part of point cloud?
Regards.
uasghar wrote:
I tested your plugin today on Photogrammetrically point cloud consisting of over 80 million points, and I must thank you that it worked better than my expectations, for very irregular surfaces (e.g. erosion gullies with very sharp spikes) it CSF performed very well. However, the settings I used were 0.2, 800, and 0.5 (with both steep slope options checked). I actually moved step by step from cloth resolution of 1 to 0.1 while keeping all other values constant, and until 0.2 my bare-earth model constantly improved. However, when I changed from 0.2 to 0.1, my bare-earth model had less points (especially in areas with erosion gullies), what I got for 0.2. Do you know if there's any specific reason behind that?
I'm trying to run CANUPO classification, but the training results show up blank. The error I get is: "Entity/DB has a null bounding box! Can't zoom in...". I've tried to run classifier training on both my own data and the scene from Dimitri's tutorial. I'm new to CC, so it is possible that I'm missing something simple. I'm following the workflow from: https://www.youtube.com/watch?v=XF41Qj4zaVg
Any hints would be much appreciated! Screenshot below.
Thanks for getting back to me. Yes, the error appears in the final step of the training process (I've made another attempt to attach the screenshot, below). I've tried zooming way in and out, and the data clouds don't ever show up. I used similar scales to those listed in the CANUPO tutorial associated with these data.
Hum that's weird! What cloud are you using? Is it one from Dimitri Lague or is it one of yours? Can you send it to me? (cloudcompare [at] danielgm.net)
That's indeed very strange: the Canupo plugin in the 2.8.1 version is broken, while it works fine in the 2.9.alpha version (and the code is the same :( ).
I'll see what I can do, but meanwhile you should use the 2.9.alpha version.
And to change the color, it's easier to set constant RGB color to each subset cloud (make sure to set the right color source in the cloud properties afterwards).
Another option is to define a custom (absolute) color scale to be used on the original cloud with all the classes.
Hi everybody,
I am using qCANUPO for my bachelor-thesis and I'm quite impressed about the first results. But as I'm not wanting to evaluate the performance of my classifiers only by visual impression, I would like to analyse the performance by calculating for example type 1 and 2 errors or the balanced accuracy (like Brodu and Lague did in their original paper about canupo).
Therefore, I need the total number of correctly classified points, and of the points which were wrong classified (compared to a manually classified result). Until now, I didn't find an easy solution to compute this in CloudCompare (the only solution I found is to merge the manually and automatically classified results and see the number of identical points by using the tool "remove duplicate points"). I'm not sure if there would be a better way to do this, so I was wondering if anyone could give me an advice (or even knows how Brodu and Lague did this)?
Thank you very much in advance,
Lena
