Agisoft Metashape
Agisoft Metashape => Bug Reports => Topic started by: c-r-o-n-o-s on September 02, 2021, 04:52:34 PM
-
Hello.
Since the new version, the display of the variance does not work for me very often, but not always.
The selection works via "Gradual Selection", but the display (red, yellow, blue) is not shown.
The icon for displaying the point cloud variance is also greyed out.
Can someone test this?
(As I said, it is not always...)
-
Hello c-r-o-n-o-s,
Point Cloud Variance view mode is only available after Optimize Camera procedure is run with "estimate tie point covariance" option enabled.
-
...of course. 8)
That was not the problem.
I think it's because of that:
I did the camera optimisation very often and always deleted more cameras.
At a certain point, however, the display simply stopped coming up.
Only when the camera alignment was completely reset did it work again.
Definitely a "small" bug.
If I can repeat it more often, I could make the project available...
-
Hello c-r-o-n-o-s,
Can you please check, if the following steps reproduce the reported problem on your side:
- optimize alignment with "tie point variance" estimation,
- switch to point cloud variance view mode,
- remove any of the aligned cameras.
-
That's how it is, that's how it happens.
But I don't think it is with every camera.
-
Ok, we'll check it and fix.
-
I have now opened the project from the previous version 1.7.0.
No variance is displayed here either.
-
Has this been solved. Point cloud variance option is not visible for me even on 1.8.1 after running cameras optimisation with the right selection to estimate tie points covariance.
I can do gradual selection but I cannot display the point cloud variance, which is a problem.
-
Hello maurosacchi,
I do not see any problems with point variance display in 1.8.2. Can you please send the project with the alignment results to support@agisoft.com, so that we could optimize it on our side and reproduce the issue?
-
My server is processing now the dense cloud. It has almost 13000 4K frames grabbed from video and 3M+ tie points now undergoing recursive optimisation. Will do once the process is done.
Only the project files though without the frames are almost 30 GB. Shall I share this with you via DropBox?
-
Hello maurosacchi,
For analysis it will be sufficient to save the project in PSZ format, keeping only the following content:
- aligned camera locations,
- markers and scalebars, if available,
- reference information in the Reference pane,
- tie points.
Other products, such as key points, depth maps, dense cloud and etc. can be removed from the project copy.
To share the project you can use common file transfer services, if you do - please share the link to support@agisoft.com. (DropBox would be fine).
-
To share the project you can use common file transfer services, if you do - please share the link to support@agisoft.com. (DropBox would be fine).
Done. I shared it to you via DropBox Transfer. Looking forward to understanding what I do wrong or why this feature does not work in my project.
-
Hello maurello,
Thank you for sharing the project. We have analyzed the provided data and have found that the alignment is quite unstable and consists of big number of poorly connected components (you can check that on Connectivity tab of the Survey Statistics dialog). In this case calculation of tie point variance leads to degenerated solution, but if you reset the alignment for all the cameras except for the biggest component and then optimize the alignment - you will be able to get the variance vectors for the tie points.
-
It was exactly in such a project that the problem arose at that time.
-
Hello maurello,
Thank you for sharing the project. We have analyzed the provided data and have found that the alignment is quite unstable and consists of big number of poorly connected components (you can check that on Connectivity tab of the Survey Statistics dialog). In this case calculation of tie point variance leads to degenerated solution, but if you reset the alignment for all the cameras except for the biggest component and then optimize the alignment - you will be able to get the variance vectors for the tie points.
You mean I should reset alignment for all camera groups except the biggest one and then optimize the cameras alignment?
If I do so I loose the overall alignment, only the biggest camera group remains aligned. That's not what I am after. I need all camera groups aligned to process the model correctly.
Did I understand it right?
-
Hello maurello,
I mean that for the project which is based on multiple poorly connected components it may be impossible to estimate the tie point variance. However, you can try to use gradual selection filters, if you need to clean up some outliers.
And you are correct, that resetting the alignment will result in some missing parts of the final model, so probably you shouldn't sacrifice the alignment in order to get variance calculation working for one or two major alignment components.
But also please note that the cameras with just a few valid tie points will not be used for the depth maps generation procedure, and if there are clusters of such cameras, you may still observe poorly reconstructed geometry in the related area or even the holes in the model.
-
I tried again. No luck. I do not get point variance in this project. What am I doing wrong?
-
I am lost with this. I tried to realign the whole project. Still do not get any point cloud variance. I do not understand how you manage to get it. Deleting camera groups won’t help me, I need them all to be able to create the whole model.
-
Hello maurello,
If there are multiple poorly connected components in the alignment (with few valid tie points connecting them) variance estimation wouldn't be possible.
-
Hello maurello,
If there are multiple poorly connected components in the alignment (with few valid tie points connecting them) variance estimation wouldn't be possible.
Clear. My lesson learnt here is that I should select the source frames better to cover the whole object. Now I tried to restart from zero and recreate frames (from video).
I come across another problem when trying to run a camera optimization for point cloud variance.
I get this error in red details window during the optmization:
Back buffer dpr of 2 doesn't match <_NSViewBackingLayer: 0x6000037e2640> contents scale of 1 - updating layer to match.
Back buffer dpr of 1 doesn't match <_NSViewBackingLayer: 0x6000037e2640> contents scale of 2 - updating layer to match.
What does it mean?