Sometimes reconstruction region is not imported

I have a CLI code which has the following part in it:

-detectMarkers %detect_marker_params% ^
-selectMaximalComponent ^
-renameSelectedComponent "Main-Component" ^
-importGroundControlPoints %marker_coordinates% %import_marker_params% ^
-update ^
-setReconstructionRegion path/to/recosntructionRegion ^

It always detects markers well and imports marker coordinates as defined in params, but sometimes it does not import the reconstruction region and instead it creates the default, initial one, which is way larger than what I want to have. Reprocessing the same scan can sometimes fix it, sometimes not.
What can cause this kind of behavior?

Hi Serob,
I tested this maybe 20 times and the region was still placed correctly for me. I adapted your script to my data.
What kind of images are you using?

I’m using 150-250 12 mpx raspberry images came from rotary body scanner. Are you testing on 1.5 version? I forgot to mention that it is happening on 1.4. And only for some scans: most of the scans are working fine, just some of them have the issue.

Yes, I tested in 1.5, as it is the current version. Can you check in that version?
Is it the similar dataset as you shared with us. In that case it could be the cause of the issue.

Yes, it is very similar and comes form the same scanner. Why that can cause the issue?

As it was causing the issues in meshing part.
The way of the capturing is not ideal for photogrammetry. From first point of view it seems like insufficient. But hard to say… Can you share the GCP’s coordinates file and the reconstruction region’s file?

Sure, where can I upload them? If that’s the same URL, I’ve lost it, can you re-share please?

https://epicgames.box.com/s/rl2i1ss8tj3lky0ed2ml4d6thurqxkrf

Thanks!
Files uploaded.

Thank you for the additional data.
I tested it and for me the region was always imported to the proper position.

Thanks a lot.
Any idea what can cause it, where can I pay attention to?

Hard to say, as I am not able to reproduce.
What is the difference between the project where it works and where doesn’t?

The curious thing is that there is no difference: same scanner, same amount of images, same environment, but one of them is working,the other one is not.

But when you check the alignment, is it the same?

Yes, didn’t find any correlation between alignment quality and this issue. Moreover, on the exact same project ot can have the issue on first run, but work as expected on the second.

Have you tried it also in 1.5 version?

Tried couple of times. Did not face today, did not face that yet.