Measure Curvature

   5292   10   0
User Avatar
Member
555 posts
Joined: Feb. 2017
Offline
Can anyone explain what curvature means ? No I'm not being silly:

http://forums.odforce.net/topic/32809-measure-curvature/?tab=comments#comment-180324 [forums.odforce.net]

Using 16.5.378, also tested 16.0.736, same problem.
Edited by vusta - Feb. 16, 2018 02:12:13

Attachments:
Curvature_Riddle.hipnc (43.3 KB)

User Avatar
Member
2120 posts
Joined: Sept. 2015
Offline
In box_2x2 the values are the diagonals of the prim faces.
In box_3x3 the values are the box face perimeter and its face diagonal x2
in box_4x4 the values are the diagonals as in from point 0 to point 7, and this x4
The other distance being the box edge x6

I guess with the node set to curvature and geometry that has straight edges, its just to going to employ some ‘random’ algorithm of which points to take for the distance.

Maybe with it set to curvature it needs to be given a ‘true’ curve and group set of points belonging to that curve in order to give some result that makes sense. I don't know haven't tested this.
User Avatar
Member
555 posts
Joined: Feb. 2017
Offline
to me it's plain wrong.

Let's pretend the corner of a box has curvature of 123.456

Now the box is 10 times as big, the same corner of that bigger box cannot magically has some different curvature than 123.456
User Avatar
Member
8766 posts
Joined: July 2007
Offline
if you multiply curvature values by the sqrt(@area) you should get stable results

not necessarily wrong, just not very intuitive, I'm sure it has logical explanation, but personally I've never used curvature values from Measure SOP

dependence on the area may make sense if you think about it in a relative terms to the world, like how sharp something is, a knife blade is super sharp, but having 1000x scaled knife would hardly be as sharp, not sure just trying to make sense of the thinking behind
Edited by tamte - Feb. 16, 2018 18:29:03
Tomas Slancik
FX Supervisor
Method Studios, NY
User Avatar
Member
4189 posts
Joined: June 2012
Offline
The definition of curvature is a ratio, therefore it's obvious that unit length will have an effect. i.e. walking on a hay bale is vastly different to walking on the earth.
User Avatar
Member
555 posts
Joined: Feb. 2017
Offline
tamte
if you multiply curvature values by the sqrt(@area) you should get stable results

not necessarily wrong, just not very intuitive, I'm sure it has logical explanation, but personally I've never used curvature values from Measure SOP

dependence on the area may make sense if you think about it in a relative terms to the world, like how sharp something is, a knife blade is super sharp, but having 1000x scaled knife would hardly be as sharp, not sure just trying to make sense of the thinking behind
1) disagree…the tip of the knife is still as sharp
2) if you have to ‘manipulate’ the answer to get what you like…that's tantamount to cheating…and proves the orig answer was wrong in the first place….otherwise, why would you tamper with the orig answer ?
Edited by vusta - Feb. 17, 2018 04:10:27
User Avatar
Member
555 posts
Joined: Feb. 2017
Offline
fuos
The definition of curvature is a ratio, therefore it's obvious that unit length will have an effect. i.e. walking on a hay bale is vastly different to walking on the earth.
yes…the earth is round but because it's huge (relative to humans, it's tiny to the Sun) it feels flat (but if you simply ‘zoom’ out…the earth is still same size but hey presto…doesn't look flat anymore does it?)

But a cube, there's corners, completely flat areas and corner edges…3 distinct areas….if their curvatures magically change simply because of size…well you're just making excuses for Houdini. I like Houdini but i'm not a fanboi who refuses to see faults in Houdini, if I think it's wrong, i'll say it's wrong…..of course, there's every chance I'm wrong.
Edited by vusta - Feb. 17, 2018 03:54:18
User Avatar
Member
4189 posts
Joined: June 2012
Offline
Definition of curvature:

k=1/r

Applied to the hay bale > earth > sun:

0.666 = 1/1.5
0.0002 = 1/6300
0.000001429 = 1/700000

Applied 1 unit box relative to its point normals:

k=1/length
k=1/1
k=1

Applied 2 unit corner

k=1/2
k=0.5

Applied 4 unit corner

k=1/4
k=0.25

therefore larger corners get smaller curvature.
User Avatar
Member
8766 posts
Joined: July 2007
Offline
vusta
tamte
if you multiply curvature values by the sqrt(@area) you should get stable results

not necessarily wrong, just not very intuitive, I'm sure it has logical explanation, but personally I've never used curvature values from Measure SOP

dependence on the area may make sense if you think about it in a relative terms to the world, like how sharp something is, a knife blade is super sharp, but having 1000x scaled knife would hardly be as sharp, not sure just trying to make sense of the thinking behind
1) disagree…the tip of the knife is still as sharp
2) if you have to ‘manipulate’ the answer to get what you like…that's tantamount to cheating…and proves the orig answer was wrong in the first place….otherwise, why would you tamper with the orig answer ?

1. Well, whether you agree or not it is one way to look at the cutvature, I personally wouldn't want to try to cut my steak with 1000x bigger knife as I don't agree it's as sharp
Already mentioned how you can make it stable
Or just make your own for the time being
Just that Measure SOP doesn't provide the type of curvature you want doesn't make it wrong, feel free to submit RFE to add more curvature types

2. Just added the analogy of the knife, no other modification to the answer, not that I care that much either
Tomas Slancik
FX Supervisor
Method Studios, NY
User Avatar
Member
555 posts
Joined: Feb. 2017
Offline
cube example, while the corner is not sharp enough to pierce your skin, if you push your finger against it it makes a certain indentation, now keep your finger there….while I magically extend the 3 imaginary lines to infinity….ie. the cube is huge…..I guarantee you the indentation on your finger is still there..(test this on the nearest corner of a table you can find).

I can see you counter-argue with, well if you put your finger on a marble, feel the curvature, now imagine the marble is enormous……like earth…it feels flat…yet is anyone going to advocate well let's go back to the idea of a flat Earth ? The Sun is enormous to Earth….but is it any less ‘round’ ?


http://forums.odforce.net/topic/32809-measure-curvature/ [forums.odforce.net]

now get the file by petz, add a transform after switch3, scale up ie. it's bigger now…colours don't change, I rest my case.
Edited by vusta - Feb. 17, 2018 23:22:27

Attachments:
Pointiness.jpg (77.0 KB)

User Avatar
Member
4189 posts
Joined: June 2012
Offline
vusta
now get the file by petz, add a transform after switch3, scale up ie. it's bigger now…colours don't change, I rest my case.


Normalize the curvature values if you want constants.

Attachments:
curvature_Norm.hipnc (277.1 KB)

  • Quick Links