I’ve had a chance to work on a bit of Augmented Reality lately, very interesting stuff. You’ve no doubt seen the many mind boggling things people have come up with this past year.  Thought I’d share a few things I  discovered.

After working with both the FLARToolkit alone and FLARManager, I prefer FLARManager. Not only is it quicker for setting up a basic AR application, the matrix smoothing on the positioning makes a big difference in the quality of the overall experience . Eric Socolofsky is actively working on the project, and every release so far has been another leap forward.

Download FLARManager here.
A good article on FLARManager by it’s author.

Another thing I noticed was that there is a definite relationship between the resolution of the marker you use and the performance of your application. That is to say, my app ran at a much higher frame rate with a 8×8 marker than it did with a 32×32 marker. This makes perfect sense after you crack open one of the .pat files that are used for marker detection. Less numbers to crunch per iteration equals faster performance.

This creates a potential dilemma. If you are building your AR app for a client, they are most likely going to want something branded for their marker, as opposed to simple black squares that work the best for detection. The solution seems to be to design a marker image that makes use of the large solid color areas that the FLARToolkit likes , but still has some nice details that will make the client happy. Then you can set the resolution 16 or 8, and let the detection ignore the details in the marker. I found it helps to design the marker on a grid that’s the same resolution you are using for detection.

A good example of this is the Disney Fantasmic Marker.

UPDATE: Another very good article, this one on custom marker creation.