Re: [Algorithms] Converting object space vertexes to bind space

[Dan T. <da...@ar...>]

I actually was mulling this over a few months ago and have a question.

You can trivially get the inverse bind pose, and the animation, and 
concatenate that *for each matrix*, but can you actually get every 
vertex in a smooth skinned model into an "untransformed" space (thats 
valid)? Been a while since linear, but I couldn't figure it out. You'd 
have to take a linear combination of the inverse bind pose and transform 
using that, but I don't think you can take a linear combination of 
matrices(certainly not with rotations!). I know I coudlnt' get it to 
work, so I just take advantage of the fact that all my animations are 
against the same skeleton and pre concatenate the inv. bind pose and the 
animated skeleton.

-Dan

Tom Forsyth wrote:
> Typically, you leave your mesh in a standard pose - sometimes called the
> "rest pose" - rather than transforming it into some sort of bind space. Many
> people use the "crucifixion" pose, though the "easy rider" pose is now
> becoming more popular as it puts some bend in all the major joints, which
> minimises stretching during animation.
>
> Obviously you know the bone orientations for your rest pose. And you know
> the bone orientations for the pose you want to be in. So you can calculate
> the delta between them - how to deform a mesh in the rest pose into the pose
> you want. This is just the inverse of the rest pose times the required pose.
>
> Then feed those deltas in as the deformations. At each vertex, blend &
> weight the deformations just like you'd expect, and apply them.
>
>
> TomF.
>
>
>
> -----Original Message-----
> From: gda...@li...
> [mailto:gda...@li...] On Behalf Of David
> Clyde
> Sent: 23 March 2007 01:34
> To: Game Development Algorithms
> Subject: [Algorithms] Converting object space vertexes to bind space
>
>
> So, this week I have been implementing a quaternion based hardware skinning
> system based on the GPU Gems 4 "Hardware Skinning with Quaternions" similar
> to the recent thread on this list with the same name. I have found lots of
> interesting information regarding the topic and various methods for
> skinning. Sadly there has been one concept this week I have been unable to
> wrap my brain around correctly. Looking for code examples online I have
> found very few with the original mesh setup included, and those used DirectX
> functions to do as such hiding the implementation.
>
> >From what I can tell the vertexes I am working with (from a Collada
> exporter) are in object space. I am currently attempting to transform the
> vertexes into a bind space which works whenever they are influenced by only
> one bone, but as soon as the vertexes are affected by more than one bone I
> seem to become lost as to what to do. My theory is that I need to accumulate
> the weighted transformations affecting the vertex similar to what is done in
> the vertex shader described in the article except using the inverse bind
> matrices for the bones then multiply it out. This isn't giving me the
> results I expected though so figured I would see if I am even on the correct
> track as I have been racking my brain over this for a while now.
>
> Thanks,
> Dave Clyde
>
>
> --
> No virus found in this incoming message.
> Checked by AVG Free Edition.
> Version: 7.5.446 / Virus Database: 268.18.12/724 - Release Date: 16/03/2007
> 12:12
>
>