Hello,
I carefully created mask for images that are active during the meshing part, so thoses masks are defining where there should be no mesh.
Here is my result (image exemple/mask/result) :
How do I fix this? Can I remove triangles thanks to masks?
It’s urgent.
Thank you all!
Ben
Hi Ben,
all of your images have created masks?
Have you captured the fence also from other side?
Have you tried Select marginal triangles option in the Advanced Mesh Model tools?
The issue there could be, that RealityCapture is creating watertight meshes. How many images have you used?
What do you have set for How to use masking layer?
Hi Otrhan,
Thank you for your reply.
I haven’t captured the other side of the fence.
I have selected the option “Only in meshing”.
I have created mask for only 25 pictures out of 200.
Select marginal triangles don’t work.
Even with the watertight meshing strategy, I don’t understand why some vertices appears as they are “blacklisted” from the mask
When I submit the post it i was only using 50 images. Now with more images it’s better but some vertices should not be there…
For such objects the other side would be almost necessary.
Also, I would use that mask for all your images, 25 is not enough I suppose.
How does look that selection? Have you tried to set it more specifically (like select large triangles option there).
This also really depends on the way, how the object is captured.
For example this is much simpler fence captured by around 300 images:
Thanks again for your fast reply
I understand the process, I was not able to cpature the other side.
Why is the mask not filtering the vertices / triangles ? Because it is it’s role…
It is possible that those vertices are from not masked images.
Mask should exclude vertices and taken in account for every vertices created…
What’s the point of making mask if they are not taken in account…?
It’s obvious as it is a manual task that they cannot be created for all images.
I guess metashape is excluding points from mask (boolean from cameras views)
Is there a way to change the weight of the mask ?
What Can I do ?
This is the extreme example, but in this case only one head was masked. As the other parts are visible from other images, which are not masked, they are basically used to mesh creation (as they are allowed):
Ok, In my opinion it’s not the definition of a mask, if it’s not taken in account.
I understand that your algoritym do not take all images during reconstruction.
Mask should be considered as strong constraints and not randomly used.
In your dropdown menu, it’s specified that it will be used for meshing. but it’s not… You should change it to “potentiel mask”…
(I already used AI for background extraction, but it leads in this case to atrefacts, I spend hours fixing them, and it’s … pointless- I’m really disapointed. Nothing personnal, just frustration)
Hi Ben,
the masks are taken in account, but as you have just 1/10 images masked those areas are reconstructed from not masked images.
I tested the owl case in other software with just one mask and the result was the same as in RealityCapture.
The mentioned dropdown menu is setting just for selected input and its mask, not for other unmasked images.
What you can try is set the Minimal distance between two vertices to a higher value in the Reconstruction settings/Mesh calculation.
I understand that your algorithm do not take all images during reconstruction. - this depends on the settings for each image. if it not set, then all aligned images are used for reconstruction.
