12 replies to this topic

[cloudfx]

Hi, I am trying to fill up a grid with circles with random radius like below rough drawing but the actual circle counts has to be about 300,000.

What would be ideal way to do this in houdini?


I've tried to use bullet to throw bunch of sphere primitives and let me fille a box and flat it to get the shape but if there is more procedural way to do this, please let me know!

[cloudfx]

cloudfx, on 30 July 2012 - 08:58 AM, said:

Hi, I am trying to fill up a grid with circles with random radius like below rough drawing but the actual circle counts has to be about 300,000.

What would be ideal way to do this in houdini?


I've tried to use bullet to throw bunch of sphere primitives and let me fille a box and flat it to get the shape but if there is more procedural way to do this, please let me know!

Attached Thumbnails

• 

[eetu]

forums.odforce.net/index.php?/topic/12892-square-packing/
forums.odforce.net/index.php?/topic/6494-poor-mans-circle-packing/

might be of help?

[cloudfx]

eetu, on 30 July 2012 - 09:18 AM, said:

forums.odforce.net/index.php?/topic/12892-square-packing/
forums.odforce.net/index.php?/topic/6494-poor-mans-circle-packing/

might be of help?

Super helpful.

Thank you eetu.

[jkunz07]

cloudfx, on 30 July 2012 - 10:34 AM, said:

Super helpful.

Thank you eetu.

I wrote a python circle packing otl a while ago, my computers at the genius bar right now but I'll try to post it tomorrow

[jkunz07]

 CirclePackOTL.zip   54.49K   74 downloads

[eetu]

Pretty shape-packing art; http://www.zeitguise...olean-taxidermy

[jkunz07]

http://mrkunz.com/ci...acking-theorem/

If your still looking for a solution, I just posted some experiments I've been doing.  Using the pcfarthest vop might be the easiest way to go if you want to place a really large # of points.

[edward]

I was thinking that you might get a "better" looking distribution of points if you ran an iteration or two of Lloyd's relaxation on it before calculating the circle radii.

Attached Files

•  LloydsRelax.hipnc   79.09K   66 downloads

[jkunz07]

edward, on 13 August 2012 - 09:19 PM, said:

I was thinking that you might get a "better" looking distribution of points if you ran an iteration or two of Lloyd's relaxation on it before calculating the circle radii.

Cool!  That's a very interesting technique - thanks for sharing it!

I was also considering using noise to color the geo before scattering points or applying a particle system with basic magnetic properties to get some intersting arrangements.

What do you think would be the best way to get all the circles to be mutually tangent?  I was thinking of using a solver sop with a foreach loop and slowly growing each circle, but I'm not sure if there's a less "brute force" method?

Edited by jkunz07, 14 August 2012 - 02:10 AM.

[edward]

You can try using closest the neighbour point after the Triangulate2D as your radius.

[edward]

edward, on 14 August 2012 - 04:21 PM, said:

You can try using closest the neighbour point after the Triangulate2D as your radius.

Er, scratch that, it doesn't work that well. Although you could try using it as your initial guess before incrementally expanding your radii.