No Pity for the Masses

A few posts back I mentioned a paper out of MIT called “Automatic Rigging and Animation of 3D Characters” and its associated code “Pinocchio.” I decided to start giving Ogre animation a run through and figured I’d post my initial work. You can find the source in my repository under “AnimationTest.” Please note that I did not include the MIT source code, which can be obtained in the links above.

The example is pretty straight forward, I hope. Not like HavokTest1. I’m using Ogre3D 1.4.9 and the Pinocchio API. Associated with the Pinocchio API is a DemoUI program that should give you a good idea on how to implement their code.

I’m not going to include a tutorial with this project. As a rule I seem to suck at writing tutorials. I’ll let the code do the talking and if there are any questions feel free to email me or post a comment. I’m fairly responsive.

:Notes – In no particular order:
These are just some things I had to deal with when integrating Pinocchio with Ogre.

1.) Pinocchio doesn’t use a global namespace. This means you either have to add one yourself or you’ll have to avoid using the Ogre namespace. These 2 APIs share a few commonly named classes. It’s a hassle. You might notice I started to use one in my code but stopped. If you download my source just remove all instances of “using namespace Pinocchio” and “Pincchio::”.

2.) Pinocchio stock mocap data does not animate in place. The example data included with DemoUI has motion. Obviously right? I don’t just mean that the person’s arms or legs move. I mean that the data physically propels itself around the environment.

3.) DemoUI gives a good idea of how to render your scene. The problem is that you are forced to update the Ogre mesh every frame tick with a new vertex array, it’s an expensive operation. Alternatively you could simply create vertex tracks for the Ogre.mesh file format or do mass memory copies.

4.) DemoUI is a great reference. Some of the mesh code I use was pulled directly from DemoUI. You can also see how DemoUI loads mocap data in the Motion class. It’s hacky at best. I’ll be rewritting this portion of the code to handle mocap data from Carnegie Mellon. The initial DemoUI requires a very specific bone format that the author does not supply. When I’ve got a free moment I’ll pull it from the code.