Leg deform

[Solkar]

About Version r5085 (io_import_scene_mhx.py, maketarget, and mh_utils copied to the resp. blender 2.66a scripts dirs)

When I follow this advice
http://makehuman.org/doc/node/corrective_shapekeys.html
I indeed can alter the shape of the leg for the "extreme"¹ pose but it seems stuck to that afterwards regardless of the bend, even if one manually creates keyframes for VisualLocRotScale parallel to the armature keyframes.

In addition, the current blender manual
http://wiki.blender.org/index.php/Doc:2 ... Shape_Keys
unfortunately has only empty pages for the related topics
Animating Shapes
Shape Keys
- Editing Shape Keys
- Animating Shape Keys
- Shape Keys Examples

¹For the record: "sitting on heels" is "extreme" only in terms of the max. rot angle of LoLeg, but not from a usage point of view - that's an everyday pose.

[Solkar]

OK, it's s "i" for "insert keyframe" while the shape slider has focus.
http://www.youtube.com/watch?v=gDZcmAWL2jA 3:20 min

But that only assigns the morph to keyframes but does not map
LoLeg X Angle -> LoLeg Morph.
once and for all

[Solkar]

...and when I try to add a driver of type "Scripted Expression" with a var related to "Rotational Difference" to that shape key Blender 2.66a simply crashes...

[Solkar]

..still no idea why that craches, but HEUREKA!

This

does the job ("kiko" being the name of the armature).

[duststorm]

Nice find.
I'm not familiar with using deforms when animating in blender myself, so I'm not able to help. But maybe Thomas does.
I hope in the future the shapekeys won't be necessary anymore, when we have a rig that takes care of it.

[Solkar]

I'm currently working on python-scripting that key creation and driving.

I'm wondering if, because of the very simple form of the backside of the human leg from just underneath buttock to just above ankles, one could try surface fitting that with a low-order polynomial and work with a bit of diff'geo and measure to get a correct volume-preserving morph.

Unfortunately I'm not overly familiar with numpy/scipy; I normally do numerics with Fortran and C++ only.
From that I know (and appreciate) http://www.ics.forth.gr/~lourakis/levmar, which is an implementation of Levenberg-Marquardt (wiki, [Mar63]) algorithm.

The core would not have to be an mh runtime module, but only a develoment tool for generating shape keys - would that also violate mh "pure python" philosophy?


[Mar63] Marquardt, D. An Algorithm for Least-Squares Estimation of Nonlinear Parameters. Journal of the Society for Industrial and Applied Mathematics, 1963, 11, 431-441. doi:10.1137/0111030

[Manuel]

https://www.makehuman.org/doc/node/advanced_some_notes_about_the_base_mesh_helpers_and_fittings.html
https://www.makehuman.org/doc/node/the_ ... panel.html

Trying to access those docs triggers a chrome SSL-error because they redirect to the site "makehuman.jwp.se".

Ah ok. The correct links are:

http://www.makehuman.org/doc/node/advan ... tings.html
http://www.makehuman.org/doc/node/the_m ... panel.html

[Solkar]

Thx, Mamuel, unfortunately the main image embedded here

http://www.makehuman.org/doc/node/advanced_some_notes_about_the_base_mesh_helpers_and_fittings.html

is not displayed.

Btw - concerning http://makehuman.org/doc/node/corrective_shapekeys.html:
I've found the 'correctives' dir for the respective targets - but where are data files for the drivers driving them?
Or are the drivers programmatically breeded up-from-scratch on export?

Regards, S.

[Solkar]

I'm wondering if, because of the very simple form of the backside of the human leg from just underneath buttock to just above ankles, one could try surface fitting that with a low-order polynomial

A polynomial fit indeed works quite well - even with a scipy-ignorant like me at the keyboard curve_fit's resistance was futile

Of, course, a polynomial ansatz is not exactly brilliant, a z-axis harmonics and elliptic xy-profile give quite a natural ansatz function and put more value in it - I stay tuned to that.

[duststorm]

Or are the drivers programmatically breeded up-from-scratch on export?

I fear that they are buried in some .py source file in shared/mhx/
Try mhx_drivers.py