An Evaluation of AR Tool Kits Among Platforms

I stumbled on this interesting article when looking for research performed with AndAR, primarily to see if there had been an expansion to support markerless AR tracking.  What I found here is more of a light evaluation of different tool kits for developing AR applications.  The authors of this paper were not performing benchmark tests or formal evaluations but rather experimenting with the feasibility of developing across several platforms and SDK's and relating their experiences and performance achieved by each SDK.  To test this, the authors designed a basic AR application that was modeled from typical archaeological surveys.  Each grid of the "site" was represented with a card or stack of cards, which a handheld device would register.  The grid squares would at the beginning each display an undisturbed patch of dirt, and deeper layers of cards may contain objects drawn in 3D.  This in addition with registered "tools" allows for a more immersive experience in simulated archaeology.

In order to accomplish this application, several base functions were designed for all the functionality of the application.  Then the AR was implemented on Android and iOS devices, using different SDK's.  For the iOS, the developers used ARToolKit for the iPhone, which is distributed commercially by ARToolworks--the company that developed ARToolKit.  They  found that while the end performance in tracking was acceptable (ARToolworks claims that this SDK can track up to 30 fps) and ARToolKit itself was usable, the team spent a significant amount of time trying to understand the iOS framework and properly implement it.  In comparison, the Android AndAR (also based off of ARToolkit) was relatively simple to set up but performed poorly.  This no doubt stems from the fact that AndAR is based off of the free ARToolKit, which has not been updated since 2007 and as such is not as advanced as ARToolworks' more recent projects, where the iPhone's implementation is more current.  Another freeware, Qualcomm AR (QCAR) was also tested and found to be not only usable, but also performed very well in their evaluations.  So while this article did not create any additional contribution to the field of AR, it is helpful to read the results achieved by developers trying different configurations.


Article:
http://www.ideals.illinois.edu/bitstream/handle/2142/27688/AR_Smart_Phone_Note_rev3.pdf?sequence=2