Display problem (holes) after distance computation

If you are allergic to bug trackers, you can post here any remarks, issues and potential bugs you encounter
Post Reply
adequat
Posts: 5
Joined: Thu Sep 29, 2016 9:54 pm

Display problem (holes) after distance computation

Post by adequat »

I have an odd display bug that results in rectangular holes of different sizes in the point cloud that partially fill and shift when moving/zooming the point cloud. This happens after loading two point clouds and computing the cloud/cloud distance (with a max. distance set, not sure whether this matters). Afterwards the point clouds show these holes (see picture). Any ideas? Thanks.
The point clouds have around 10 to 20 mio points each and are displayed alright before computing the distance. When zooming out enough the holes vanish. However, zooming in again and they reappear.
CC_DisplayBug_1.jpg
CC_DisplayBug_1.jpg (54.39 KiB) Viewed 3230 times
daniel
Site Admin
Posts: 7332
Joined: Wed Oct 13, 2010 7:34 am
Location: Grenoble, France
Contact:

Re: Display problem (holes) after distance computation

Post by daniel »

Hum, that looks like a LOD bug?! Can you try to deactivate the LOD display? You have to uncheck the 'Decimate clouds over XXX points when moved' option in the 'Display options' dialog.

If this works, then I'll be interested in knowing:
- if this problem only occurs after computing the distances (or if you already have it before when zooming in)
- does the problem persist if you re-activate the 'Decimate clouds...' option but disable the 'Try to load clouds on GPU for faster display' option
- what is your graphic card model?
Daniel, CloudCompare admin
adequat
Posts: 5
Joined: Thu Sep 29, 2016 9:54 pm

Re: Display problem (holes) after distance computation

Post by adequat »

Thanks for the quick help (and this great piece of software, by the way). Switching off the "Decimate clouds over ... points when moved" does indeed fix the problem.
But there is some weird behavior:
Before the comparison I have not observed the holes.
When first trying switching off the option, the cloud with the scalar field from the comparison did show OK but the other one without the SF did not show at all. Switching the option off again made both clouds visible again but also brought the holes back. However, when running more trials later with two simpler clouds, both clouds appeared just fine after switching the option off (and with holes when switching the option on again).
Also, applying the Tools -> Cleaning -> Noise filter on one cloud created the _cleaned cloud which worked OK even though the original showed holes (I found this before I read your answer so the decimation option was active).

Running the comparison with the "Decimate ..." option turned off and turning it on only afterwards seems to work fine. Both clouds do not show holes. (EDIT: this only seems to be true for smaller PCs of around 10 to 20Mio points, when switching back to larger ones with >50 Mio the holes appeared also in this case)

The "Try to load clouds on GPU..." option does not seem to have any effect. Increasing the cloud size for the "Decimate clouds" option to above the size of my clouds does, again, fix the problem (this setting is somehow adjusted from 10Mio to 50Mio when playing with the GPU option, which confused me a bit at first).

Currently I use a Nvidia GTX1070 on Windows but I had very similar artifacts on my laptop with an Intel graphics unit (also Windows).
I hope this helps.
daniel
Site Admin
Posts: 7332
Joined: Wed Oct 13, 2010 7:34 am
Location: Grenoble, France
Contact:

Re: Display problem (holes) after distance computation

Post by daniel »

Thank you very much for the detail feedback. I'll try to understand what's at stake here.

it's probably a bug with the LOD display. Hope I'll be able to reproduce it...
Daniel, CloudCompare admin
daniel
Site Admin
Posts: 7332
Joined: Wed Oct 13, 2010 7:34 am
Location: Grenoble, France
Contact:

Re: Display problem (holes) after distance computation

Post by daniel »

Okay, I finally found the bug! It was occurring when computing distances (or using the ICP algorithm which also computes distances). The LOD structure was not updated while the octree could be modified.

I've uploaded a new 2.9.alpha version online.

And thanks for the feedback ;)
Daniel, CloudCompare admin
adequat
Posts: 5
Joined: Thu Sep 29, 2016 9:54 pm

Re: Display problem (holes) after distance computation

Post by adequat »

I just updated to the latest version and indeed the holes are gone. Perfect, thanks a lot!
Post Reply