Note: This is the second installment of a two-part conversation with Marxent Software Engineer Ken Moser, PhD. In the first part of the interview, Ken explained what Markerless Augmented Reality is, the technology behind it and the ways it’s being used in the real world. Click here to read Markerless Augmented Reality: A Q&A with Marxent’s Ken Moser, PhD, Part 1.
Q: What sets Marxent’s Markerless AR system apart from these other systems?
Ken Moser: Marxent’s proprietary Markerless technology uses algorithms that lie right in the sweet spot between the static off-line and complex SLAM Markerless approaches. Our system works on-line with live video from a mobile device and is able to track the user in any environment provided it has a floor. However, our solution does not build a map of the surroundings. It instead provides an iterative update of the user’s position relative to wherever they started. This “Relative Tracking” yields a far more immersive result compared to static off-line alternative, but does not require the initialization, calibration or mapping overhead of SLAM. This makes it ideal for visualization of virtual content within small spaces and applications running on current mobile hardware.
What was the process of developing it like?
Development of any tracking solution is never trivial, to say the least. There are enumerable design considerations and constraints that must be balanced, not the least of which are performance and accuracy. If the algorithm is allotted the maximum amount of computation time needed to return the most accurate pose estimation, the performance of the application, in terms of rendering frames per second (FPS), will likely be extremely low and may even be unusable in real-time settings. Of course, relaxing the pose estimation threshold for speed will decrease the tracking accuracy over time and the system will exhibit some jitter and drift during use. We’re constantly tweaking constraints and tuning formulas in an effort to provide users the smoothest experience possible.
You must be very proud of the way Marxent’s Markerless AR turned out?
In terms of more personal aspects, being a part of developing Marxent’s AR tracking solutions has been incredibly fulfilling. It is always rewarding to be given an opportunity to push your own intellectual limits, as well as work on projects that will provide literally millions of people a unique, never before seen experience. In a previous article I noted that I performed some of my PhD research at the Interactive Media Design Lab at the Nara Institute of Science & Technology in Japan. The lead professor, and founder, of this lab is Dr. Hirokazu Kato, one of the two co-creators of ARToolkit, the first and most widely used open source marker tracking solution, now overseen by DAQRI. I am very proud to be able to apply many of the lessons I learned at NAIST to continue the advancement of tracking for consumer AR.
So how long until we’re all using Markerless AR on a daily basis?
I think the only real hurdle right now for ubiquitous adoption of these tracking solutions is accessibility, in terms of ownership and pricing, to the technology. ARToolkit was first introduced in 2001, and still remains the only viable open source option for marker-based tracking, with the alternatives being pricey subscription-based licensing from Vuforia and similar purveyors. Markerless tech has, so far, begun the slide to a similar fate, with small start-ups developing SLAM solutions being quickly bought up by large companies like Apple and SnapChat, relegating the solutions to either unused software stockpiles or nonsensical social media plugins. Hopefully though, the availability of Markerless offerings will avoid, to a much greater degree, the suppression seen in marker tracking and similar solutions, letting those small or solo development teams with truly novel ideas to have access to the tools they need to change the world.