Disney gave a talk at Siggraph this year about their implementation of DQ. I haven't checked, but if you're lucky the talk material might be up on the ACM library.
http://s2013.siggraph.org/attendees/talks/events/enhanced-dual-quaternion-skinning-production-use On Tuesday, November 12, 2013, Sergio Mucino wrote: > Thanks Raffaele! We'll keep investigating. Cheers! > > > On 11/11/2013 11:48 PM, Raffaele Fragapane wrote: > > By default DQ doesn't consider scaling. The original paper omits it too as > it's not that common at bone anim level in games (which is what the paper > focused on). > > We had the same issue with the DQ compounds when they were offered and > we tested them, so we ended up cooking our own to account for scaling. It's > not that hard to do, but it's more than just adding scaling up or > downstream from DQ unaltered. > > As it's work done for Animal I can't disclose details, sorry about that, > but if it's worth anything to you to know: it's perfectly possible with no > noticeable effect on performance and not particularly hard to do. > > > On Tue, Nov 12, 2013 at 8:16 AM, Sergio Mucino > <[email protected]<javascript:_e({}, 'cvml', > '[email protected]');> > > wrote: > >> I have an interesting issue at hand. It seems that the shipping Dual >> Quaternion Deformation ICE compound does not take into account scaling. I'm >> using it for one of our rigs, and scaling down the rig scales things >> properly, but the deformations on the mesh are still being calculated in a >> non-scaled space, it seems (they are quite exaggerated when the rig is >> scaled down). >> First, I thought I could try fixing it myself by multiplying the >> calculated deformations by the scale factor from the rig, but that didn't >> work. I googled a bit, and found a Dual Quaternion Deformations 2 compound, >> which seems to support scaling. I downloaded it, but it made no difference. >> Taking a closer look at it, it's doing exactly what we were doing here to >> fix the first compound (taking the matrix from the mesh, inverting it, and >> multiplying it by the vector coming from the skinning nodes), but it made >> no difference. The deformations are still not scaling down. >> I turned off the ICE tree and re-enabled my envelope (it's still there, I >> just mute it when enabling the ICE tree), and the deformations are ok there >> (I just need to use DQ on this mesh, that's why I'm not using the >> envelope), but re-enabling the ICE DQ deformer gives me the wrong >> deformations. >> Has anyone seen this before? Any known cures for this ailment? Thanks! >> -- >> > > > > -- > Our users will know fear and cower before our software! Ship it! Ship it > and let them flee like the dogs they are! > > -- Technical Director @ DreamWorks Animation
<<Sergio Mucino_Signature_email.gif>>
<<image/gif>>
