[Soc-2011-dev] Weekly progress report 7 - Improving Motion Capture Workflow

[Benjy Cook]

This week:
- Constraint system is fully in place, with support for baking/unbaking. 3
Constraints have been implemented: Distance, Freeze Point, and Maintain
Point. This allows users to easily fix common issues with post retargeting
artifacts.
- Related to constraint baking, management of layer system has been coded
(leveraging the NLA system), where on the bottom you have the retargeting
motion, then the constraints that are baked. A "tweak" layer will be added
to this, allowing the user to manually fix small issues easily.
- Creation of "original motion" empties will now only be created if the user
chooses IK for the retarget, and the relevant bone does not have an IK
constraint. I still prefer creating empties over new bones, as new bones
cannot be created in proxy skeletons.
- Some cosmetics to the UI. Also, now each bone has a small button to mark
it as a "foot" bone or not, meaning that the retargeting process should
create the translation of the root by looking at these bones and ensuring
that they do not skate. Without marking these bones, translation is not
transferred to the user root bone.
- Added 2 new preprocessing tools:
- A simple "Fix Armature Rotate" which addresses the common issue of mocap
files having a rotated axis system to Blender's.
- Added denoising operator. Uses a type of median filter to smooth out
spikes, typical of sensor noise in motion capture data.
- Some bugfixing to accept and catch some more types of hierarchies, and
some other small issues with the stride bone system.

Next week:
- Modify existing code to leverage new PoseBone functionality (thanks to
Campbell!), where now we can directly edit the world matrix (in some cases).
Also some code refactoring and more extensive commenting - while everything
works there are some unoptimized/messy code sections.
- Create a more extensive tutorial, covering the constraint system as well.
- Start extensive testing of various mocap and rig sources, to ensure there
are no issues there.
- Floor constraint still needs to be coded
Also in the upcoming week or the next:
- Path editing - enable parenting of stride bone to a curve, and making sure
that the parameterization matches the original translation rate (i.e.
setting the evaluation time curve correctly).
- Code some hierarchy matching heuristics to ease user's mapping experience.


Issues: None.

Benjy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/soc-2011-dev/attachments/20110708/cfa00f61/attachment.htm