Measure Curvature

[Noobini]

an oddity, or could someone please explain why ?

I have 3 boxes, take the 1st one which consists of just 8 points, all at the corners..we can see they all have curvature of 1.41421 (in spreadsheet)

Next box has 1 more segment, so we have some different curvatures for the new points...but where are the 1.41421 points...they can't just suddenly have different curvatures now ? They are STILL at the corners.

Then next box has more segments....curvatures change again....this is simply illogical, for points lying in the 'middle' (a simple one has 4 flat neighbours) I would expect it to have curvature 0...these seem to be the only consistent ones...the rest...they change with 'resolution' which makes no sense to me.

EDIT: this takes the cake, if I increase the size of the 1st box, leaving same resolution, the curvature change !!!....I guess i have no idea what curvature means now...

(seems inversely proportional to edgelength ?)

Curvature_Riddle.hipnc

Edited February 16, 2018 by Noobini

[tamagochy]

To get real curvature i think better to convert box to vdb and use vdb analysis.

How work measure sop i have no idea))

Results very different between this two methods.

[f1480187]

Curvature usually works nice on dense and smooth meshes. There is no general method to measure curvature on all kinds of inputs. The one on Measure node probably has to do something with averaged point normals on the neighbours. It also scale-dependent, so, if you scale box by number of divisions (hence maintaining size of quads) it will stay at 1.41421. It is a square root of 2, which has nothing to with the problem, but I still want to point it out.

 

I recommend @petz uploads on curvature.

Various mesh curvature algorithms, including mean and gaussian curvatures, also principal curvature directions:

Same but using VDBs:

 

[Noobini]

so obviously my plain English understanding of the word 'curvature' is completely different to the scientific meaning of curvature. It's a bit like asking what's the angle of a square's corner ? 90 degs....doesn't matter what size/rez whatever that square is, the corners are always 90degs. If someone (Measure SOP) tells me the 'curvature' of a cube's corner is different to that same corner of a bigger cube...oh well, as suspected, I don't know what curvature means. Thanks for all input.

Edited February 16, 2018 by Noobini

[Arthur Spooner]

Would the VDB method work for vellum simulations to get the curvature of cloth ? I have some cloth which is making strong creases which I want to soften.