Equivalent to Maya's Wrap Deformer...

[Sergio Mucino]

I have a mesh that I need to have deform by another. I cannot do a GATOR weight transfer because the source mesh is being driven by an ICE tree reading and setting point positions, not by weights. I was wondering what would be the best way to achieve this. I did a little bit of research, and it seems that the Cage Deformer is regarded as somewhat slow. Any other options? Can ICE do something similar? Thanks for any help!

--

[Sergio Mucino]

My other problem is that in some parts, this mesh I'm trying to fit overlaps other meshes, so It needs to be deformer by more than one source mesh.

[Sergio Mucino]

Okay. I can see why the Cage deformer is regarded as slow... daaaaaamnnn!
That's definitely not gonna work for us (plus, there's a weird offset coming from somewhere). I'm gonna have to cluster these meshes over to the mesh they're trying to follow... ugh...

[Mirko Jankovic]

What about Shrink Wrap? Hmm it is also for 1 mesh only I think...

[Steven Caron]

it is slow to bind, but useable after you freeze the weights.

i would look into an ICE solution, store the locations in modeling stack, create a deformer in animation, and freeze the modeling stack

[David Gallagher]

Really, I've had much better luck with CageDeform than Maya's WrapDeformer, which tends to be unstable and poppy.

Have you tried different settings here?

Is there an ICE CageDeform out there?

[Manny Papamanos]

I use cage all the time, it has saved my rear quite a few times and I can't complain about its performance.
Like David said, reducing the 'falloff' helps immensely, (those cage settings are sometimes overlooked cause they are pretty deep).. Plus, be sure that your cage is as low poly as possible, this makes a huge difference.
Run your cage object through a poly reduction before the cage if needed.

Manny Papamanos
Product Support Specialist
Softimage and Motionbuilder

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of David Gallagher
Sent: Thursday, October 24, 2013 4:36 PM
To: soft...@listproc.autodesk.com
Subject: Re: Equivalent to Maya's Wrap Deformer...


Really, I've had much better luck with CageDeform than Maya's WrapDeformer, which tends to be unstable and poppy.

Have you tried different settings here?

Is there an ICE CageDeform out there?

[cid:image0...@01CED0D8.1A091800]

On 10/24/2013 2:05 PM, Sergio Mucino wrote:
Okay. I can see why the Cage deformer is regarded as slow... daaaaaamnnn!
That's definitely not gonna work for us (plus, there's a weird offset coming from somewhere). I'm gonna have to cluster these meshes over to the mesh they're trying to follow... ugh...

[cid:image0...@01CED0D8.1A091800]

On 24/10/2013 3:45 PM, Sergio Mucino wrote:
My other problem is that in some parts, this mesh I'm trying to fit overlaps other meshes, so It needs to be deformer by more than one source mesh.

[cid:image0...@01CED0D8.1A091800]

On 24/10/2013 3:42 PM, Sergio Mucino wrote:
I have a mesh that I need to have deform by another. I cannot do a GATOR weight transfer because the source mesh is being driven by an ICE tree reading and setting point positions, not by weights. I was wondering what would be the best way to achieve this. I did a little bit of research, and it seems that the Cage Deformer is regarded as somewhat slow. Any other options? Can ICE do something similar? Thanks for any help!
--

[cid:image0...@01CED0D8.1A091800]

[Sergio Mucino]

I'm not sure exactly what you're talking about Steve... :-)
Sorry... Softimage noob in training here...

[Steven Caron]

sorry to tease... start by trying the 'deform by hull' compound.

[Sergio Mucino]

Cage deformer is definitely not going to work. For some reason, the moment I add it, the mesh it was added to (the driven mesh) is being shifted to the side. I can't have that offset (I'm not even sure why it is being created).

Is there a way via ICE to get the point positions of one mesh, and set them on another mesh that has a different topology? Basically, I want to look at the closest point from the source mesh, and set its position on each point of the target mesh (the deltas will be 0, since the positions are exactly the same... the target mesh is a copy of the source mesh, but with several polygons deleted).

[Grahame Fuller]

I don't think that Deform by Hull will be any faster than Cage. I believe that Cage is multithreaded and pretty optimized already (it shares code with the envelope deform). The real issue with speed is that, since it binds to polygons or vertices, there can be a huge amount of deformers in the list. Try to optimize the cage mesh as Manny suggested, and adjust the strength and falloff as David suggested (this will limit the number of deformers in total, and the number per vertex).

I would also try freezing the weights as Steven suggested, otherwise they are live and potentially being recalculated. In an explorer, expand object > Polygon Mesh > Clusters > CageWeightCls, select Envelope_Weights, and click Freeze. How slow is it now?

gray

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Steven Caron
Sent: Thursday, October 24, 2013 5:03 PM
To: soft...@listproc.autodesk.com
Subject: Re: Equivalent to Maya's Wrap Deformer...

sorry to tease... start by trying the 'deform by hull' compound.

On Thu, Oct 24, 2013 at 1:56 PM, Sergio Mucino <sergio...@modusfx.com<mailto:sergio...@modusfx.com>> wrote:
I'm not sure exactly what you're talking about Steve... :-)
Sorry... Softimage noob in training here...

[cid:image0...@01CED0DB.F1F67840]

On 24/10/2013 4:29 PM, Steven Caron wrote:
it is slow to bind, but useable after you freeze the weights.

i would look into an ICE solution, store the locations in modeling stack, create a deformer in animation, and freeze the modeling stack

On Thu, Oct 24, 2013 at 1:05 PM, Sergio Mucino <sergio...@modusfx.com<mailto:sergio...@modusfx.com>> wrote:
Okay. I can see why the Cage deformer is regarded as slow... daaaaaamnnn!
That's definitely not gonna work for us (plus, there's a weird offset coming from somewhere). I'm gonna have to cluster these meshes over to the mesh they're trying to follow... ugh...

[cid:image0...@01CED0DB.F1F67840]

On 24/10/2013 3:45 PM, Sergio Mucino wrote:
My other problem is that in some parts, this mesh I'm trying to fit overlaps other meshes, so It needs to be deformer by more than one source mesh.

[cid:image0...@01CED0DB.F1F67840]

On 24/10/2013 3:42 PM, Sergio Mucino wrote:
I have a mesh that I need to have deform by another. I cannot do a GATOR weight transfer because the source mesh is being driven by an ICE tree reading and setting point positions, not by weights. I was wondering what would be the best way to achieve this. I did a little bit of research, and it seems that the Cage Deformer is regarded as somewhat slow. Any other options? Can ICE do something similar? Thanks for any help!
--

[cid:image0...@01CED0DB.F1F67840]

[Grahame Fuller]

That's weird - is there a Static_Kinestate property already on the object before you add it as a cage?

gray

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Sergio Mucino
Sent: Thursday, October 24, 2013 5:06 PM
To: soft...@listproc.autodesk.com
Subject: Re: Equivalent to Maya's Wrap Deformer...

Cage deformer is definitely not going to work. For some reason, the moment I add it, the mesh it was added to (the driven mesh) is being shifted to the side. I can't have that offset (I'm not even sure why it is being created).

Is there a way via ICE to get the point positions of one mesh, and set them on another mesh that has a different topology? Basically, I want to look at the closest point from the source mesh, and set its position on each point of the target mesh (the deltas will be 0, since the positions are exactly the same... the target mesh is a copy of the source mesh, but with several polygons deleted).

[cid:image0...@01CED0DC.D2C82E40]

On 24/10/2013 4:43 PM, Manny Papamanos wrote:

I use cage all the time, it has saved my rear quite a few times and I can't complain about its performance.

Like David said, reducing the 'falloff' helps immensely, (those cage settings are sometimes overlooked cause they are pretty deep).. Plus, be sure that your cage is as low poly as possible, this makes a huge difference.

Run your cage object through a poly reduction before the cage if needed.



Manny Papamanos

Product Support Specialist

Softimage and Motionbuilder

From: softimag...@listproc.autodesk.com<mailto:softimag...@listproc.autodesk.com> [mailto:softimag...@listproc.autodesk.com] On Behalf Of David Gallagher

Sent: Thursday, October 24, 2013 4:36 PM

To: soft...@listproc.autodesk.com<mailto:soft...@listproc.autodesk.com>

[Sergio Mucino]

Thanks a lot Steve. This looks like could do the trick. However, I'm having some problems with it. It seems my target mesh is missing something, because when I try to use the compound, three of its internal nodes are in an error state.
Get Self.deformerpointLocation: is not initialized
Get Self.NormalPerpendicularness: is not initialized
Get Self.OffsetDistance: is not initialized
I only plugged a Get Data node set to "self" into its Cage Geometry input... I guess I'm missing something so the un-initialized nodes can get the data they need, but I'm not really sure what... any help is welcome. Thanks again!

[Sergio Mucino]

There is. All scale components are set to 1, and all Rot/Pos components to 0.
What is this for? I don't think I had seen it before...

[Sergio Mucino]

Sorry, I meant I picked the mesh to use as a Cage in the Get Data node

[Steven Caron]

well the most straight forward implementation in ICE will have a lot less data. not a list that is number of verts squared but just the number of verts.

[Grahame Fuller]

It's used for a few things, but mostly to keep track of the bind pose so that Softimage knows when a deformer has moved. If it's there it means that at some point you used that object as a deformer. If you're not currently using it as a deformer then you can safely get rid of it, then try to add the object as a cage deformer again. A new property will get created with the current transform values.

gray

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Sergio Mucino
Sent: Thursday, October 24, 2013 5:22 PM
To: soft...@listproc.autodesk.com
Subject: Re: Equivalent to Maya's Wrap Deformer...

There is. All scale components are set to 1, and all Rot/Pos components to 0.
What is this for? I don't think I had seen it before...

[cid:image0...@01CED0DF.75CDA960]

[Grahame Fuller]

Deformer or deformee, I should say.

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Grahame Fuller
Sent: Thursday, October 24, 2013 5:36 PM
To: soft...@listproc.autodesk.com

Subject: RE: Equivalent to Maya's Wrap Deformer...

It's used for a few things, but mostly to keep track of the bind pose so that Softimage knows when a deformer has moved. If it's there it means that at some point you used that object as a deformer. If you're not currently using it as a deformer then you can safely get rid of it, then try to add the object as a cage deformer again. A new property will get created with the current transform values.

gray

From: softimag...@listproc.autodesk.com<mailto:softimag...@listproc.autodesk.com> [mailto:softimag...@listproc.autodesk.com] On Behalf Of Sergio Mucino
Sent: Thursday, October 24, 2013 5:22 PM
To: soft...@listproc.autodesk.com<mailto:soft...@listproc.autodesk.com>
Subject: Re: Equivalent to Maya's Wrap Deformer...

There is. All scale components are set to 1, and all Rot/Pos components to 0.
What is this for? I don't think I had seen it before...

[cid:image0...@01CED0E0.3D5A1A90]

[Sergio Mucino]

Looks like I will have a few problems here...

1. I cannot reduce the cage polygon count. Actually, it is a hi-res model (not crazy polygon counts, but still). I need to attach to it a copy of itself, but with lots of polygons deleted. Kinda like attaching polygon patches that need to follow it. So, I'm trying to get a lower-res mesh to follow the high-res

2. The high-res mesh is deforming. It actually has its points set to follow another mesh (another copy) via ICE. The original mesh is the one with all the deformation operators on it (pretty complicated stack of deformations). We don't touch this third mesh... that's why this second high-res has a get/set point positions ICE tree. It's the render mesh.

I'm thinking of a different solution... hold on... I might have something that works...

[Sergio Mucino]

Okay. Say I have two identical meshes (as I've explained before). Mesh A is the original... Mesh B is a copy of mesh A, but it has missing polys (imagine it has holes in it).
How could I select on Mesh A the polys that are left on Mesh B?

[Grahame Fuller]

Oh I must have missed that, sorry. On the original object, select the polygons that you don't want to delete and choose Model > Create > Poly Mesh > Extract Polygons (keep). The selected polygons will be copied to a new object (the old ones will be left alone). As long as you don't freeze the Extract Polygons operator, it should update properly.

gray

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Sergio Mucino
Sent: Thursday, October 24, 2013 5:45 PM
To: soft...@listproc.autodesk.com
Subject: Re: Equivalent to Maya's Wrap Deformer...

Looks like I will have a few problems here...

1. I cannot reduce the cage polygon count. Actually, it is a hi-res model (not crazy polygon counts, but still). I need to attach to it a copy of itself, but with lots of polygons deleted. Kinda like attaching polygon patches that need to follow it. So, I'm trying to get a lower-res mesh to follow the high-res

2. The high-res mesh is deforming. It actually has its points set to follow another mesh (another copy) via ICE. The original mesh is the one with all the deformation operators on it (pretty complicated stack of deformations). We don't touch this third mesh... that's why this second high-res has a get/set point positions ICE tree. It's the render mesh.

I'm thinking of a different solution... hold on... I might have something that works...

[cid:image0...@01CED0E1.ECFD10F0]

On 24/10/2013 5:35 PM, Grahame Fuller wrote:

It's used for a few things, but mostly to keep track of the bind pose so that Softimage knows when a deformer has moved. If it's there it means that at some point you used that object as a deformer. If you're not currently using it as a deformer then you can safely get rid of it, then try to add the object as a cage deformer again. A new property will get created with the current transform values.



gray

From: softimag...@listproc.autodesk.com<mailto:softimag...@listproc.autodesk.com> [mailto:softimag...@listproc.autodesk.com] On Behalf Of Sergio Mucino

Sent: Thursday, October 24, 2013 5:22 PM

To: soft...@listproc.autodesk.com<mailto:soft...@listproc.autodesk.com>

[Matt Lind]

If the construction history on Mesh B is still live, then the indices of the polygons should be the same as on Mesh A.  Otherwise  you’ll have to resort to using metadata (if already applied) or do something like a get closest location with the meshes aligned the same global pose.

 

Matt

[Sergio Mucino]

Ah, if I had known that... :-)
I spent like half an hour deleting polys on my copy. But of course, I ended up with a completely independent object. Sound like what you suggest would retain a connection to the original object, which is what I'd need.
I will give it a go.

[Sergio Mucino]

That's what I thought, but it doesn't seem to be the case. I retained the history because of this (I can see a DeleteComponents Op folder full of all the operators used to delete the polys I wanted to get rid of), but if I select all the polys left over, create a cluster with it, and move it to the source mesh, the polys selected by the copied cluster are totally different polygons to the ones I stored in the cluster, which told me that the order changed. I'm kinda lost on this one...

[Sergio Mucino]

This is definitely gonna work. Thanks Grahame! I'll continue tomorrow.

[Sergio Mucino]

Reviving this topic...
I got back to working on this issue, and found a couple of issues.
First, I'll describe what's going on. I've got a character's coat, to which we're attaching some mud patches. To do this, I selected the polys under the mud patches, and extracted them as suggested before. Everything works (the extracted mesh does follow the original coat's deformations). My idea was to use these extracted polys as a cage for a cage deformer on the mud patches.
I'm having two issues. First, creating the cage deformer on the mud patches mesh, using the extracted mesh as the cage, has thrown out any sense of interactivity out the window, big time. The rig went from real time to "wait 10 seconds for anything to happen". The animators will definitely not work with this.
Second, I don't know if it's related to the above or not, but it seems the cage deformer is not really working (as expected, at least). There's a mud splat on the forearm, for example. When SI finished applying the cage deformer to the mud patches, I can see the weight points all over the mud mesh. Fine. However when I rotate the forearm up (and after waiting for a while for anything to happen), the patch on the forearm loses all its weight colors and goes completely blue, staying in its place, instead of following the forearm. Its as if the cage that influences it walked away, leaving it with no weights (instead of it following the cage). I don't know if its because the deformer does not understand the procedural deformations happening on the extracted mesh or what. I also tried moving the cage operator from the Modeling stack over to the Animation stack. It made no difference.
In any case, the complete grind in performance may stop me from using this approach at all. I'm open to additional suggestions. If anyone knows of alternative ways (ICE?) of deforming a mesh with another, I'm all ears.
Before anyone suggests this one... GATOR is not an option (the original coat uses a very complex system of layered deformations involving envelopes, lattices, and ICE trees. GATOR seems to ignore a lot of these things).
Cheers!!

[Steven Caron]

did you freeze the weights? the performance should pick up once you freeze the weights. also, use settings like where mentioned earlier in this thread (search distances and such)

also, i don't know much about your pipeline but why does animation need to see these mud patches? surely this could be a job for character effects department? or maybe they are hidden right before point caching?

lastly, an ICE cage deformer is possible, here is one way to do it...

http://download.autodesk.com/global/docs/softimage2013/en_us/userguide/index.html?url=files/idef_deforms.htm,topicNumber=d30e295689

not the most ideal because of the mesh copy required, which can that can be removed if one knows the cavaets with ICE attributes, ref models, and ICE optimizations

[David Barosin]

My guess is that when you extract polygons it makes a new object with the center at world 0,0,0.   The extracted polygons are not really deforming with the source geo but rather getting extracted at each frame in the new pose.  So the cage deformer is seeing this as a new bind pose rather than a deformation.   

[Sergio Mucino]

You're right. Freezing the weights did help. However, for whatever reason, the mesh is still not picking up the cage's deformations. It just stays still... no idea what's going on here.
I'll look into the ICE alternative. I'll see what I can come up with.
I'll also check with the CG sup about having these patches bypass the anim dept. We cache everything afterwards anyway, but I need them to move first.
Thanks Steven!

[Sergio Mucino]

Ah, that makes sense. And unfortunately for me, throws this solution out the window. Time to look at something else. Thanks David!

[Jeremie Passerin]

Hi Sergio, 

I wrote a simple mesh deformer with ICE, that I'm using to replace the cage deform in some cases. 

To use it, you need to create two ICETree. 

1. The first one is a initialization. Give it the mesh that you use to deform, and two ICEAttribute to store the original state.

2. In the second one just plug the previous ICE datas.

3. FREEZE THE MODELING (that's deleting the first ICE Tree)

And that's it.

Compounds are attached to this email.

[Sergio Mucino]

Thanks a lot Jeremie! I'll give it a shot (I tried using the ICE Deform by Hull compound approach, but it's not working, very likely due to what David explained). Cheers!

[Steven Caron]

hey jeremie

how is this working with referenced models? and if you duplicate a character (same reference duplicated in a scene)?

when i was at blur we made an ICECageDeformer which relied on the freezing technique but had some troubles in the areas above. either ICE dropping the attribute due to optimizations, or if it couldn't find the correctly named object in the scene, or in the case of the duplicated character all the duplicates would follow only the first one created in the scene.

@sergio, test this in your pipeline THOROUGHLY!

[Steven Caron]

you might be able to skip the 'extraction' step because ICE might be faster.

[Sergio Mucino]

Well, for some reason, this one isn't working either. I've tried referencing different coat meshes (my extracted mesh, the one I extracted it from, which basically has a get points pos/set point pos ICE tree on it referencing another mesh, and the deforming coat the second one references). Nothing. My patches still don't move.
Looks like I'm a sitting duck on this one. I'm gonna end up doing it the old fashion way... enveloping the mud patches to cluster centers on polys on the extracted mesh. Not pretty, but at least, I expect it to work...
Thanks a lot everyone!

[Jeremie Passerin]

Hi Steven, 

Well this is something I have created recently so maybe I haven't spotted the issues you're talking about. 

Though, I remember Enoch mentioning that the ICECageDeform wasn't working in 2010 but does in 2013. So maybe something has been fixed. 

My version use a similar approach, I simplified a little bit and added blend and scale support.

Jeremie

[Manny Papamanos]

Don't mean to hijack.
>From my tests on an envelope using extract then using those extracted polys as a cage, won't work.
Use the actual skin mesh instead or duplicate it and delete what you don't need and use that as a cage.

The solution I would use is to GATOR the mud to the char.
Merge all the mud if you want to have a single element.
Gator that to the character.
Freeze the GATOR op.

Gator is mostly practical because you will be dealing with 'weight deform' once frozen which is very fast.


Also, you can save alot of time by testing on a simple scenario first and not on the production sample.

Manny Papamanos
Product Support Specialist
Softimage and Motionbuilder

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Jeremie Passerin
Sent: Tuesday, October 29, 2013 4:11 PM
To: softimage
Subject: Re: Equivalent to Maya's Wrap Deformer...

Hi Steven,

Well this is something I have created recently so maybe I haven't spotted the issues you're talking about.
Though, I remember Enoch mentioning that the ICECageDeform wasn't working in 2010 but does in 2013. So maybe something has been fixed.

My version use a similar approach, I simplified a little bit and added blend and scale support.

Jeremie

On 29 October 2013 12:55, Steven Caron <car...@gmail.com<mailto:car...@gmail.com>> wrote:
you might be able to skip the 'extraction' step because ICE might be faster.

[Sergio Mucino]

Thanks Manny. Earlier in the thread I explained why GATOR won't work for me in this case. Thanks anyway!

[Manny Papamanos]

Don't know if you have the option but you can use Plot>shape on the point cached mesh .
I then tested Cage and that worked, GATOR also worked even when frozen .

-manny

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Sergio Mucino
Sent: Tuesday, October 29, 2013 4:32 PM
To: soft...@listproc.autodesk.com
Subject: Re: Equivalent to Maya's Wrap Deformer...

Thanks Manny. Earlier in the thread I explained why GATOR won't work for me in this case. Thanks anyway!

[cid:image0...@01CED4D1.2F309710]

On 29/10/2013 4:24 PM, Manny Papamanos wrote:

Don't mean to hijack.

>From my tests on an envelope using extract then using those extracted polys as a cage, won't work.

Use the actual skin mesh instead or duplicate it and delete what you don't need and use that as a cage.



The solution I would use is to GATOR the mud to the char.

Merge all the mud if you want to have a single element.

Gator that to the character.

Freeze the GATOR op.



Gator is mostly practical because you will be dealing with 'weight deform' once frozen which is very fast.





Also, you can save alot of time by testing on a simple scenario first and not on the production sample.







Manny Papamanos

Product Support Specialist

Softimage and Motionbuilder

[Leonard Koch]

If I'm understanding you correctly, then LK Skinner should solve your problem.
You can get it on http://www.rray.de and http://www.leonardkoch.com