Hacker News Reader

Comment by kefka | Hacker News Reader

Hmm. I thought they canceled this project.

When I demoed it months ago, it was so not ready for primetime. The AR area was pathetic, and it used IR lasers for scanning.

The AR area is self-explanatory: when the size is not much bigger than a postage stamp, your project makes a great tech demo, and a really shitty customer product.

The IR laser is problematic in multiple ways:

1. If it's the same module as found on the kinect, it's a 90mW laser, which is a hell of a lot of power on a battery platform. Expect much reduced power in scanning mode.

2. IR means the device is useless outdoors and near windows. Think of anyplace where the Kinect fails: so too will this.

I'd pay $1000 for this, not the $3k they're asking. The tech just isn't there. I just hope this doesn't cause an AR desolation in the community the way VR had done for 10 years.

View on HN

> I'd pay $1000 for this, not the $3k they're asking. The tech just isn't there. I just hope this doesn't cause an AR desolation in the community the way VR had done for 10 years. reply

And so the high price works exactly as intended. Microsoft KNOWS that this isn't ready for prime-time, so they put this price tag on it to make sure that it isn't buyed by someone just enthusiastic who is then dismayed when it is "not ready for prime-time" (as advertised!). They've stated multiple times that this is the reason for the price tag.

Cost is only a part of the problems with this platform.

IR scanning is inherently erroneous. I simply can't be:

1. Outside 2. Near windows 3. Near IR flood sources 4. Reflective/glass areas

It just doesn't work on any of those areas. That's the huge unsolvable problem with this platform. And no, frequency shifting to a different wavelength still has the same problems (the sun is a great wideband emitter).

Even if their platform was brought to the best their hardware could bear, it still would fail.

EDIT: For those of you giving -1's, how about you comment and explain why you are doing so. I'm providing a technical argument why this platform will fail, based on my experiences with this headset and my work in computer vision.

"That's the huge unsolvable problem with this platform"

Unsolvable problems tend to be unsolvable. I can solve this with three words: Use it inside. And on the next version (which, let's be honest, they're targeting a "next version" for general release, not this one), they use lasers instead of IR.

If you want to use it outside, it might be unsolvable for the current release, but it's hardly unsolvable for the platform.

The IR emitted is a structured laser light at IR frequencies.

They're already using "lasers". The problem with using any sort of structured light is that your frequency can't be in the same band of any sort of flood lights (sunlight, ballasts, IR camera lights). It's also why X-Rays work well: it's a structured light (kind of) and can be used because we won't have a flood of XRays everywhere.

The problem isn't with "IR": they should not be using a structured light solution. It's just not durable enough for 3d scanning applications. It's also why I've worked on a SLAM-like 3d scanner. Understanding geometry, corners, edges, and gradients are much more versatile in scanning, because it is passive. Our very eyes don't send out beams of light, and yet we manage a very accurate SLAM-like map inside our own heads.

The problem with the Hololens is that until they switch to a vision-only system, they're going to be plagued with problems regarding IR sources (or whatever freq they choose).

Source: me (3Dollar scanner, amateur radio operator, OpenCV reprap implementer)

"Use it inside" doesn't completely solve it, at least if you have windows inside.

During my master internship at INRIA, a coworker had to put a tablecloth on his desk to be able to get a proper depth image with a Kinect. It turned out to be a small window, 5 meters away, reflecting on his desk. Indirect sun light was enough to mess up the projected pattern.

So "Use it in a cave" can be a solution, but it's... less attractive.

Funny fact is, Hololens doesn't work with windows.

90mW (+ ~90mW waste heat) is 180mW. WiFi & LTE each draw about 700mW.

Reply on news.ycombinator.com