hmmm … ok i think u will need to review the whole thing (most probably) from the start;
retargeting is essentially :
original (source) data/character/motion → Retarget Filter (RF) → new (target) output/data/character/motion
we tried methods which we hoped were going to quickly fix your problem, clearly not; that means u have a bigger issue;
you need to rethink what your RF must do and under what circumstances:
- what ratio you have
- how to apply the ratio (positions and/or rotations)
- which bones need to be affected and which not
- which time frame to apply these on
- very often in RFs, end-effectors may also be introduced
- inverse kinematics can be used to resolve some constraints
- dynamic hints can be introduced
- maybe your matrices need matrix Jacobian interventions, etc…
- and more…
u could look for papers (some are paid, others are free) which have resolved similar issues such as this one ( http://graphics.snu.ac.kr/publications.php look for the word “Retarget” ) and see how u can make yours work or behave in the same fashion as in a paper…
About the source data:
- review your source data thoroughly and find any info u could have missed about it
- or you may want to try different source data such as the likes of http://mocap.cs.cmu.edu/ or mixamo
- or work on a smaller skeleton that has the same problem (easier to debug)
this will give you confidence that your RF is behaving as expected;
anyway, i hope u can make progress, maybe take a short break from it and come back refreshed to retackle it;
until then ?
