Agisoft Metashape
Agisoft Metashape => General => Topic started by: JJ on June 06, 2021, 07:17:51 PM
-
i'd like to build my meshes from depth maps as it saves a lot of time, but the resulting geometry is extrapolated to the point of becoming unusable.
as i missing something or is this a limitation?
see attached showing both meshes. the mesh from dense cloud also has a texture hence the additional sharpness but it's the meshing that i'm questioning.
-
any views on this?
those of you building meshes from depth maps only, are you getting these extrapolations?
-
Hello JJ,
Which settings you are using for the both approaches (parameters in Build Dense Cloud and Build Mesh dialog)?
I think that the extrapolated parts could be filtered out and removed by the confidence filter for the depth maps based mesh generation approach.
-
thanks Alexey - please see attached.
with the dense reconstruction i have always used Default interpolation .
with the depth maps i tried Default and Disabled but the result is the same.
-
I think that the extrapolated parts could be filtered out and removed by the confidence filter for the depth maps based mesh generation approach.
how would i go about filtering the extrapolated parts out?
currently looking to build a different mesh and have the same problem.
-
I think that the extrapolated parts could be filtered out and removed by the confidence filter for the depth maps based mesh generation approach.
i found the Model COnfidence view but i still can't find how to filter the Depth Maps.
AFAICT you can only filter by confidence the Dense Cloud right? in which case the speed advantage of generating the mesh without a dense cloud is lost
-
AFAICT you can only filter by confidence the Dense Cloud right? in which case the speed advantage of generating the mesh without a dense cloud is lost
here's a screenshot of where i am now. i can view the model confidence but not able to select by confidence like i can with the dense cloud
-
I think that the extrapolated parts could be filtered out and removed by the confidence filter for the depth maps based mesh generation approach.
forgive me for insisting Alexey but you hinted that it could be possible to filter the extrapolated geometry (which would save me so much time /and/ improve the models) but have left me no clues as to how to achieve this
i have Metashape Standard - is it possible to filter out the extrapolated parts without building the dense cloud?
anyone?
thanks
-
extrapolated triangles are normally of bigger size. You can try to select triangles (gradual selection) by size and see if it helps up to certain point.
-
thanks but i already tried that and there's just no consistency in the parts that get removed
pity i can't get this to work because apart from the extrapolated parts the generated mesh is generally better defined and it's a much quicker process, but there's simply too much made up stuff
the bit i'm struggling to understand is that Alexey says the extrapolated parts can be removed with the confidence filter for the depth maps but i only see confidence filtering for the sparse cloud, dense cloud and mesh, not for depth maps
am i missing something simple here?
-
I've found this very unrefined and dirty way to filter the mesh by confidence using meshlab...
If you export your model as PLY then you get the option to also export vertex confidence values. Make sure you also uncheck binary encoding.
There's probably a more elegant way to do this, but if you open your exported PLY file in a text editor you'll see a list of vertex property names near the top, with 'confidence' being the last one. Replace the word 'confidence' with the word 'quality' and save the file.
Open the file in meshlab, and select Filters -> Selection -> Conditional Vertex Selection.
In the 'boolean function' text box type something like (q < 5) and hit preview.
q = quality (confidence) so fiddle with the number until you are happy with which vertices have been selected. You have to tab away from the text box to update the preview.
Press Apply
Press Close
Select Filter -> Selection -> Delete selected vertices
They are gone!
-
thanks for the suggestion James. i went through the process but using polygon size as a filter doesn't work as the parts i need to delete don't necessarily have larger polys.
so AFAICT it's either mesh from Depth Maps which is better defined mesh but unwanted extrapolated polys or mesh from dense cloud which will give a better cropped mesh, though not as defined.
-
Sorry, it hadn't occurred to me that polygon size would be so closely correlated with vertex quality, but I guess it makes sense in a way.