Over the past few months I've been working on and off on a generator for creating arbitrary complex fiducial markers for use in reacTIVision[1] projects. Having been under extreme time pressure for our Lexus installation[2] at Detroit Autoshow I initially designed a system based on a shape vocabulary. This was a promising start but had some shortcomings since about 10-20% of the markers were not recognized if they're part of a bigger set (though all unique)...

The new generator is creating markers much closer to the original set published by the reacTIVision developers. However you can set various parameters to control the overall complexity and size of the markers generated and so optimize the design for different usage scenarios. In the above example, I allow markers to become quite large and all trees are made of 30 nodes. The layout is physics based using the same Verlet physics engine as in this softbody demo[3]. Btw. This Verlet engine will be part of the next release of toxiclibs and is already available via SVN[4].

The generator itself will be released under GPL in the next couple of weeks.

[1] reactable.iua.upf.edu/?software
[2] youtube.com/watch?v=WEU6_O2jE10
[3] vimeo.com/702576
[4] code.google.com/p/toxiclibs/source/browse/trunk/toxiclibs/src/toxi/physics