My fascination with AR

About nine months ago I developed this intense fascination with Augmented Reality. I even wrote about it at one point. Well that fascination is still with me, so I thought I’d describe about what I’ve discovered in that time.

The first thing I found is that many AR advocates and researchers are pursuing things that seem a little pointless to me. For instance, this ad for the mini lets you see a 3D mini model by waving a magazine around in front of your webcam. It would be much easier to use if they just put a flash app on their site with a slider that let you spin the car. Then there are the games that you can only see through a cell phone. It’s like a regular 3D game only with unreliable flickering graphics and bad UI. The ad is at least commercially viable for the ad agency, if not for Mini. Nobody is going to buy the cell phone invisible train game.

“But that’s not the point!” you might be thinking.  To me commercial viability is exactly the point. The best way to be able to keep working on thing X is to make it worth buying. You sell v1.0 so you can use the money to make v2.0. If v1.0 is worthless even to early adopters, and you can’t trick some VC into giving you millions of dollars for no reason, you don’t get to make v2.0. I’m a fan of having a business model and I’m not afraid to admit it.

The second thing I’ve learned about AR is how close we are to having all the pieces while still remaining frustratingly far away. Computing power is cheap enough to fit a wearable in a relatively small bag. GPS receivers are less than $30, and that’s not in quantity. Cameras that are good enough to do basic feature tracking from are under a hundred bucks. Between WiMAX and 3G phone networks, mobile internet connections are widely available. That is most of the pieces of a real first person AR system available right now for under a couple hundred bucks. These guys are demonstrating a solution to many of the hard problems running in real time on a laptop.

Aside from integration, the biggest remaining hurdle for first person AR is displays. The two display vendors I’m keeping my eye on are Lumus and Microvision. Lumus was at CES in January showing off their glasses, but they are still not available for purchase. Up until August or so their website said “Fall 2008″, but that’s since switched to “Mid 2009″. Do they really have an impending product launch, or is it vaporware that will fall prey to the Econopolypse? Who knows, but I’m going to see if they’ll sell me a pair of glasses in 6 months. Microvision is working with the military on some wearable display stuff, and are looking for partners for their consumer products.

The third thing I’ve learned about the state of Augmented Reality is how little of the work is going into applications. It’s like the PC in the 1970s. Everybody working on AR starts by grabbing a computer vision library (of which there are about a dozen) and then putting together their own motion tracking or fiducial detection system. After a couple of years they get things working pretty well and then finish grad school and move on to something else. There is no standardized AR API to write against and no one has even started working on whatever the killer app for AR will be.

This last challenge is where I’ve been focusing my attention. Unfortunately the first step there is to get a motion tracking and scene mapping system of my own working. So I’ve been reading a lot about computer vision and boning up on linear algebra so I can actually understand what I’m reading. This is a part time personal project for me, so it’s not incredibly fast, but I am making progress. At this point I have no idea if this will lead me to start an open source project, found a company, or even throw up my hands and give up.  It’s been a very interesting learning experience so far though and I’m still very excited. I feel like I’m watching the early stages of a huge new industry. Hopefully I’ll have some worthwhile progress of my own to share before long.

Anyway, I’ll now return you to your very occasional reading about the game industry. :)

~Joe


One Response to “My fascination with AR”

  1. Alex Kasper commented on :

    During my studies I did a tiny bit of AR myself. I made a little software that enabled you to use your hand as a pointing device in 3D. You’d hold your thumb and index finger in a pistol like fashion and point on stuff, then bend your thumb for clicking.

    There is kind of a API for AR, called ARToolkit, maybe you’ve heard of it (http://www.hitl.washington.edu/artoolkit/)?

Leave a Reply