Delta Mush ICE

722 views
Skip to first unread message

Miquel Campos

unread,
Sep 7, 2014, 12:14:46 AM9/7/14
to soft...@listproc.autodesk.com
Hello,

I would like to share with you my Delta Mush ICE compounds (Or what I guess delta mush is :P). I am pretty sure is not the best implementation of Delta Mush (due my null maths skills ) but it is working kind of OK.  (Very slow that is true  )

This version is using a simple average neighbours smoothing (I think this is a uniform Laplacian), without any weighted average. That means that can cause smoothing artifacts if the mesh have big differences between the smallest and biggest  polygons. This artifacts normally show up with higher mush iterations.

One little trick that I found is apply a relax before apply Delta Mush.  So this uniform a little the size of the polygons

Also I found Delta mush is very impressive when you only have 1 deformer x Point. is like Voodoo magic ;)

Regarding experiments and errors, If the smoothing at origin is different that the smoothing in the final, you can control how much the detail is washout or exaggerate. I call it the body-builder effect.


I hope you like it! 

if you have any improvement or smoothing  algorithm, please share it :)

Cheers,
Miquel 

PS: you need to activate the delta mush in the sample scene. Check the Second ICE tree in the animation stack.
PS2: Eddy share your OCD compounds, we want it!! ;)


----------------------------------------------------

Miquel Campos
www.miquelTD.com

DeltaMush_ICE.jpg

Sebastien Sterling

unread,
Sep 7, 2014, 9:13:25 AM9/7/14
to soft...@listproc.autodesk.com
Looks cool ! is this all in Ice ? or does it require another mesh to be smoothed out ?

Miquel Campos

unread,
Sep 7, 2014, 9:19:30 AM9/7/14
to soft...@listproc.autodesk.com
100% ICE


----------------------------------------------------

Miquel Campos
www.miquelTD.com

Jason S

unread,
Sep 7, 2014, 4:29:11 PM9/7/14
to soft...@listproc.autodesk.com
What I dont understand is, how does it smooth-out the visible yanks in the mesh, but while not also shoothing-out the nose for example? does it use the base static mesh shape or?

nevertheless, 100% nice! :)

Raffaele Fragapane

unread,
Sep 7, 2014, 8:20:24 PM9/7/14
to soft...@listproc.autodesk.com
Delta mush isn't a smooth di per se', it's a binding state trick. 
You smooth a mesh and save the delta of points from full detail to smooth in the bind pose, you then deform and smooth out the full mesh, and lastly re-apply the deltas back on the deformed smooth mesh to reinstate the high frequency detail and irregularities.

A basic implementation like that is quite simple actually, and that's why it's kinda brilliant.

It's pretty easy to do with OOTB XSI actually if you use ICE to apply shapes after the envelope, or use ICE to skin things in the modelling stack.

Take mesh, smooth it, shape animate from smooth to HR with shapes set to local (it will use the local reference frame of each point).
Then apply your skinning, and apply a smooth deformer on top. Lastly apply the local shape through ICE to recomputed point framesets.
The reference frameset is trivial, Y is the normal, X or Z is the first edge projected on the normal's plane, the last axis is the cross product of the two.
You also have ICE facilities for that, I believe, and if they match the same system used by local shapes they should just work.

If you instead do the skinning in ICE in the modelling stack, then you have even less work to do, as you can use local shapes OOTB (they are meant to preserve local transforms in shape with something live going on in the modelling stack).
 
--
Our users will know fear and cower before our software! Ship it! Ship it and let them flee like the dogs they are!

Chris Gardner

unread,
Sep 7, 2014, 8:25:50 PM9/7/14
to soft...@listproc.autodesk.com
I prototyped a delta mush in ICE to figure out my splice version (http://www.chris-g.net/2014/09/05/delta-mush-in-splice/) just cos I wasn't that familiar with fabric. 

Disturbingly easy!

Miquel Campos

unread,
Sep 7, 2014, 8:45:04 PM9/7/14
to soft...@listproc.autodesk.com
Oh! Woah! Thanks for sharing it Chris! :)

And thnks Raffaele for the explanation.  That is what i did with the 2 nodes.


--

----------------------------------------------------

Miquel Campos
www.miquelTD.com


Raffaele Fragapane

unread,
Sep 7, 2014, 8:48:45 PM9/7/14
to soft...@listproc.autodesk.com
Given smoothing in ICE is always going to be fairly slow, I imagine if you wanted some performance out of it using the DQ skinning compound, or even just building up your own linear skinning, in ICE in the modelling stack and using local shapes and the native smooth for the mush would be by far the fastest option before taking on it with full fledged C++ custom nodes.
Unless you're only using ICE to re-apply the shape and not doing any smoothing inside it, in which case you're probably on par.

Juhani Karlsson

unread,
Sep 8, 2014, 6:58:04 AM9/8/14
to soft...@listproc.autodesk.com
Can you Raff explain what you mean with store delta of points? What does delta stand for?
Its kinda new term for me. Maybe I`ll check the compound out.. ; )
--
-- 
Juhani Karlsson
3D Artist/TD

Talvi Digital Oy
Pursimiehenkatu 29-31 b 2krs.
00150 Helsinki
+358 443443088
juhani....@talvi.fi
www.vimeo.com/talvi

Alok Gandhi

unread,
Sep 8, 2014, 7:08:53 AM9/8/14
to soft...@listproc.autodesk.com
Let me take the liberty to explain before Raff does. Delta of points mean difference between two position. Let's say that you have a point in a mesh at [x1, y1, z1] position. And then for the same mesh you apply some changes through deformers / ops as a result this point now moves to a new position [x2, y2, z2]. Then a Delta of Positions between these two positions of the mesh will be - [x2-x1, y2-y1, z2-z1]. The Delta Positions would be an array of such difference positions of all the points in the mesh. That's mathematical definition. In simpler terms, it is just the information which records the difference in the position of each vertex in a mesh between two different states.
--

Juhani Karlsson

unread,
Sep 8, 2014, 7:16:02 AM9/8/14
to soft...@listproc.autodesk.com
Thanks Alok!
Your explanation makes sense.
I had a hunch that it was something like that, but its better to ask than live in confusion! : )
Reply all
Reply to author
Forward
0 new messages