QuadRemesh - trying to fit a MeshPatch but settings do almost nothing

Rhino
1

Hi,

I have a quite simple mesh that fits through a collection of points.
I’d like to QuadRemesh this to improve the topology.
I’m fully aware that there is not way the resulting mesh is going to pass exactly through the reference points.
However, I’m quite amazed by the fact that bumping up the amount of faces doesn’t seem to do anything.

I’m using both Grasshopper and Rhino to test these settings and I get :

  • max deviation constant, about 15cm for a quad face number between 300 and 3000.
  • then it decreases to 8 cm when reaching 10000 quad faces, which is already too big for further point editing.
  • and then I tried 100000 and 1 million faces and I still get 8cm maximum deviation !

I can see more faces but they don’t get any closer to the input mesh.

See here 3000 and 100000 yet similar shape.

image
image1091×610 35.2 KB

image
image1103×626 40.9 KB

The adaptative percentage doesn’t seem to do anything either. Does it only work for Breps ?

image
image1136×630 42.3 KB

image
image1127×638 37.2 KB

I was able to fit a surface from points in Grasshopper within 3cm max deviation with only 40*40 points… That would be 1600 quads. :man_facepalming:

Mesh.3dm (194.9 KB)

2

turn off adaptive quad count- you’ll get this-

that removes any auto decimation and gives you just the specified quad count. (@Trav please correct me if i’m explaining this incorrectly)

image
image1714×1372 215 KB

3

Hi Kyle,

It doesn’t seem to make the quad mesh any closer to the input mesh.
I checked and with or without it, the number of faces is always close to the target that I set.

image
image1067×620 102 KB

image
image1090×631 79.7 KB

4

subdivide your input mesh a few times, then quadremesh, then run point deviation on it to see the real differences. I’m seeing good results here-

image
image1920×1179 196 KB

5

Thanks for trying this. Your results seem indeed quite good but I can"t reproduce those.

I ran SubDivide 7 times - mesh with 49152 faces.
Then QuadRemesh with 10000 faces, 100% adaptative size and adaptative count disabled.
Then I run PointDeviation using the original mesh as point provider and the squad as target and I see this monstrosity.

image
image1263×622 93.6 KB

Same in Rhino WIP.

A few tests with less quads and/or less subdivisions give similar deviations.

Here with 3 subdivisions (from the screenshot it ressembles yours) and 2000 quads.

image
image1253×485 174 KB

Edit : same results on another PC (W10) and a Mac.
If you remember them could you give the exact settings you used so I can try to reproduce that ?

6

Hi,

I got to test this a little more.
I think your last picture shows a lot of points that are ignored because of the proximity angle value set to 1 by default. If you bump it to 10, it shows all the points.
I get exactly the same results (12.7 cm deviation for 1000 quads, 13.2 for 3000 quads, 8.8 for 100000+ quads) as testing the distance in Grasshopper.

The initial question remains, why adding so many faces doesn’t allow the quad mesh to fit any closer to the initial mesh ? Is there a stiffness value that could be changed, similar to the Patch command ?

7

@theoutside Adaptive allows QuadRemesh to add more polygons in areas of curvature that exceed the specified target count.