Woz & Jobs 2.0: Leap Motion's Holtz & Buckwald

theodp writes "Over at Popular Science, Tom Foste takes a look at the $79 Leap Motion controller and inventors David Holz and Michael Buckwald, best friends since they were fifth graders in Florida. Potential applications for the device are many, as proof-of-concept demos ranging from controlling Windows 8 (video) to driving JPL's Athlete Rover (video) show. 'If we're successful and build something that is a fundamentally better way to interact with a computer, there are essentially an unlimited number of use cases,' Buckwald says. 'Eventually, anything that has a computer could be controlled with it—every laptop, every desktop, every smartphone, every tablet, every TV, every surgical station, every robot, potentially even a Leap in every car.' And even if 'it's got some growing pains to experience,' writes Ars Technica's Lee Hutchinson, 'it's cool-it's extremely cool. It's not yet a game-changing interface device, but it could be.'"

Comments

Re:So then..

[jklovanc]

Will Buckwald steal from Holtz too?

More PR hype, what, Leap isn't selling?

[janoc]

We have two Leaps at the office, I had a chance to play with it and look at the SDK. Unfortunately, as the recent CNN article points out, it is a solution looking for a problem.
The device is nothing else but two cellphone cameras with an USB interface and 3 infrared LEDs behind an IR filter. It tracks the infrared reflections off your fingertips (or a pencil or whatever) in 3D using stereoscopic vision. It does work, as much as that technology allows (nothing really revolutionary there), but the device and mainly its software have some serious issues:

• The device is way too small - the consequence is that the cameras are too close together and thus the tracked working volume is tiny. If you use both hands at the same time, you can barely move before you run out of space and the camera stops tracking your hand.
• It is very sensitive to occlusions - if the fingers (or some reflection) aren't visible, no tracking. E.g. making a grabbing gesture is really hard, because closing the hand into a fist hides the fingertips and the software gets confused. Rotating a closed hand is outright impossible - the software doesn't know how to distinguish the top and the bottom of the hand unless it can see the fingers (i.e. you are holding the palm open and spread out)
• The API is more targeted towards emulating a mouse or a multitouch desktop than actual 3D interaction. Unfortunately, a real mouse or a proper touchscreen are way more comfortable, accurate and robust to use for such tasks than the Leap, with its glitchy tracking and buggy software. There is no way to get the raw video from the SDK, so no custom algorithms are possible.
• The available software is crap. There is no way around it. The very buggy "Airspace store" is full of paid apps that are often little more than a buggy demo that you would throw away after 5 minutes once the novelty of wiggling your fingers at the screen wears off. Yay for drawing with your finger in the air ... Many of the apps are things that could be hacked together in Flash in a few minutes. Furthermore, many of the available apps try to do things that would be way better handled using a simple mouse or some other controller, thus there is little benefit from using the Leap.

So all in all - unless you are the type of person that wants to show off at the next Powerpoint presentation by changing slides by waving one's hands (and be a laughing stock when the device won't work or skip several slides instead), there isn't much to be excited about. It is really a solution looking for a problem.