Motion retargeting

Using Cinder
1

I have a project where I would need to apply skeletons acquired from motion capture on a 3d model.
Setting the root joint position and copying all the rotations work kind of ok, but this does not work when the original skeleton of the model and the motion captured person have different sizes. There are errors like the 3d model flying above the floor, legs are skating, etc.

To solve these issues one needs to use motion retargeting as far as I know, changing the orientations based on bone size difference, or using inverse kinematics, etc. Because this seems to be a very broad topic, maybe it is better to ask here before I dive into this if anyone has experience with this.
I would be grateful for any pointers. Thanks.

-Gabor

1 Like
2

I am also interested in this. :slight_smile: I will probably have to do something similar pretty soon.

3

Are you doing a playback after the capture has finished? I’ve had success with MotionBuilder in the past for retargeting. I think Blender may have some addons too, but I’ve never tried them.

4

@paul.houx, drat, I was counting on you :).

@clint.h, thanks. The ideal solution should work in real-time during capture. We worked with a Blender addon, which is called makewalk, and solves the issue nicely, although it can only be used for post-processing recorded data.