'parameterization__texturing_from_registered_rasters' generates black texture about pymeshlab

[alemuntoni]

Could you please share also the input model and the rasters?

[ramazanabd]

I attached the rasters, model and cameras.out

stairs_demo.zip

[MLucient]

Hi @alemuntoni - like many other folks, I want to join in appreciating PyMeshLab - it's a huge win for those of us trying to make ambitious reconstruction pipelines!

But also, I do want to +1 this issue; my team just started trying to port some things to PyMeshLab and we've been unable to fix this issue on our side so far. One suggestion/idea? I know that standalone Meshlab sometimes has some implicit rules about pathing/etc with respect to where it expects images/rasters to be, and I've seen a version of this issue (black/null texture) with this particular filter fixed by simply putting every resource in a single directory before doing any of the operations in MeshLab, so I wonder if something similar is happening here?

Also, if there's anything you'd like me to help test, I'd be happy to help out as well, just let me know!

[alemuntoni]

I am sorry that the fix of this is taking a long time, but I still didn't have time to investigate. I'll try to check this as soon as possible, but right now I am pushing for a new meshlab release.. I'll write here whenever I'll have some news.

About what is causing this trouble, I don't think that it is related to relative/absolute paths of textures... I already fixed that issue in a previous commit. I think that this is still related to the GL context, that is required by the parameterization__texturing_from_registered_rasters filter...

[alemuntoni]

This issue should be fixed with the last commit. The issue was actually related to the double precision of meshlab (2020.12d was producing black textures), and was affecting both import_cameras_for_active_rasters_from_file and parameterization__texturing_from_registered_rasters.

@ramazanabd @MLucient if you want to try the nightly that fixes this, you can download the artifact package produced by github actions (select the right one depending on your platform and your python version) and extract the content of the zip file in a directory called pymeshlab. Then, from outside that folder, use pymeshlab like you normally do. If you could try and check if it works fine also for you, it would be really appreciated!!

Note: as stated in #51, the output of compute_normals_for_point_sets in 2020.12d is different from the one computed with meshlab 2020.12.

[ramazanabd]

It works for me as well. Thanks for the fix!

[MLucient]

Yep, works for me too. Thanks @alemuntoni !