I have processed one dataset of photos taken by a tablet.
I am using WebODM and I have noticed that the point cloud I get with ODM 2.8.6 is sparser than the one I got with previous versions using the same processing parameters:
end-with: odm_georeferencing, feature-quality: medium, feature-type: hahog, pc-filter: 0, pc-quality: ultra, pc-sample: 0.035, resize-to: -1
This is the point cloud I got with previous ODM versions →
1.laz - Google Drive
This is what I get now (same parameters, same Cloud Compare version, same PC) →
2.laz - Google Drive
Any ideas about this issue? If it can help, I had a similar issue some months ago that has been reported and fixed
Thank you for your support!
Update: I tried with pc-sample=0 and the results are quite similar to the older ones. Could be the pc-sample unit of measure that have been changed in the last version of ODM?
I don’t think we have changed pc-sample behavior in newer releases, but perhaps something did get fixed and now pc-sampling is actually working properly. (0.03 means you’ll get a point every ~3cm).
Thank you pierotofy. In fact I measured the distance and I got 1 mm for the old point cloud and 35mm for the newer one. So probably it was something in the older version of ODM…
Old point cloud
New point cloud
It looks like the behavior is now correct (based this on your examples dataset, in fact). This is the pull request with the change:
Patch pc resolution by smathermather · Pull Request #1435 · OpenDroneMap/ODM · GitHub
However I see some “holes” if I process with pc-sample=0.001. I expected a point every 1mm but in some zones this seems to be different:
This is the result with pc-sample=0
The holes are probably missing matches. Is it mostly every 1mm with an occasional missing one?
31 July 2022 16:11
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.