Fool me, this is real good shit ! [the VR thread]

[Darkflame]

Oh, it will be possible eventually. I am highly sceptical on their timeframe though. Its going to be like the evolution of computer graphics. We are really just getting out of the "labs" stage and onto, say, the tracking 3d registration equivalent of TNT2 cards. Another decade and it will be fantastic.
I also dont think solving AR and gestures and voice controll is wise. Yes, they work together in theory, but its just making the mountain harder to climb. I'm happy with a 3d stylus, or even a controller for now.

What interests me more at the moment is how it works from a developer standpoint; Does the OS handle the tracking/overlay and leaving developers free to make the app? If the OS handles the actual AR does that mean we can have separate 3rd party apps in the same field of view at the same time?
I hope thats the case - as (being in a lot of AR circles) I have long been campaigning for this approach. The "do everything in the one app" approach so far is very limiting. Not to mention pirvacy implications (routing everything though the same company's servers)

The Wired hands on impressions is here;
http://www.wired.com/2015/01/microsoft-hands-on/

[pogo]

The particular thing I’m skeptical of is its ability to keep things that it’s drawing so steady, relative to the world seen through the transparent bits of the device. Problem is there is bound to be lag: in the motion detection, in calculating and rendering, and in transmitting to the screens. So as soon as you move your head it will judder. Accurate head tracking is hard. The latest Rift prototype is covered in LEDs which your PC tracks with a web cam. This has nothing like that, that I can see, so should be much worse than Rift.

It’s very ambitious; not just a sub-phone PC like Glass, more like a full PC. All in one unit so no wires. Semi-transparent screens so it can be worn all the time, not just when confined to a small space as you can’t see. It’s what I imagined Rift and Glass would evolve into; VR for the masses, not just for games. Problem is I thought and still we’re years from seeing something like this in practice.

[Absolute_Zero]

If this helps... eye tracking :O



Gizmodo hands-on.
http://gizmodo.com/project-hololens-han ... 1680934585

[pogo]

It’s what they call “pinning” that I’m skeptical of. Where you move around and it captures that information, does all the calculations, renders it and outputs it to the device so quickly that it keeps up perfectly. They can do the calculations very quickly but there the limitations of the input device and the renderer are hard to eliminate.

This normally doesn’t matter with e.g. a game as the lag is small and players adapt and compensate. If e.g. you know there’s a small lag in a driving game you steer a little earlier. The lag is much smaller than the lag due to network latency, so is minimal. Where it is an issue it’s an issue of game balance, so you e.g. make opponents even slower, or give them other handicaps.

But it will be hard to fudge when the results are drawn alongside the input. Any lag will be apparent, as will any imperfections in the head tracking. It will need to track both head orientation and position continuously and precisely; I don't see how without cameras looking at it, i.e. the Rift solution.

[gumboot]

You may (very, very optimistically) be able to pin your rendered content to the eye as a surrogate means of pinning it to the real world.

Disregarding, briefly, the possibility that the wearer is focusing on something artificial or something that is moving; the human eye should give you a fix on a point in the real world and all you have to do is keep your latest rendering square with that. The iris seems to me to be a pretty good fiducial for this job, and it's conveniently linked to a fundamental limit of human perception.

Then you have to figure out if they're watching something that is moving, or if they're watching something that isn't really there. That's when you'll have to look at your high-latency filter; but perhaps you can afford some delay at that point. Delays may imply some trippy effects, but that could come down to a case of "prove me wrong" fringe effects.

[Darkflame]

Remember our eyes/brain also have limits. Pinning (Im more used to the term registration) during large head motions can be less accurate provided they can blur correctly - so knowing the speed and direction of the turn becomes more important for that brief period then precise environmental tracking.

Not that it isnt still a huge challenge to get all this right. But sometimes its a question of managing the imperfections rather then trying to remove them completely.

[Yak]

I'm kind of surprised they've gone so heavily for this from out of nowhere. Maybe they felt they needed to have some kind of response to Oculus so we see this now but won't get hands on for years, or maybe I'm wrong and they really do have something imminent.

Until there's one on my desk next to my dk2 though I reserve judgement.

Interesting though. VR, AR, it's all interesting and the more people trying the more likely it is to break through and become something for everyone.

[gumboot]

Suppose a human fixes on an object, and eye-tracking fixes on the pupil and renders a raptor accordingly. Their head may move around a bit, but their eye compensates and eye-tracking picks this up and corrects the position of the raptor to compensate (low-latency hacks a la time warping in OR).

Where I imagine it goes wrong is when the thing the human is staring at begins to move. The pupil tracks. The eye-tracking tracks. The fucking raptor sets off with it! Shit! Moving raptor! Panic!

We'd need to recognise that pupil movement isn't necessarily compensating for head movement, but our measurement of head movement is pretty shaky, and it could take a some time to put the raptor back where it belongs. But that may just amount to a small twitch, or it may be imperceptible, or hard to quantify. I've no idea.

[gravy]

This is amazingly interesting and kind of unlike Microsoft.

Usually it's "Look at this tech, it's amazing! You could probably use it for something!" and no actual decent application of the tech.

This time they potentially have the first VR / AR killer app in Minecraft.

This is all on the assumption that the core tech works well and has very low latency. Which is a big assumption, I know


EDIT: and here's the Ars hands on too http://arstechnica.com/gadgets/2015/01/ ... tual-real/

[gravy]

It’s what they call “pinning” that I’m skeptical of. Where you move around and it captures that information, does all the calculations, renders it and outputs it to the device so quickly that it keeps up perfectly. They can do the calculations very quickly but there the limitations of the input device and the renderer are hard to eliminate.

The pinning could work in a stateful way if you assume the environment you're projecting onto isn't going to change much (ie tables and walls don't move)

So an initial scan could have higher latency but after that orientation ends up being a much simpler pattern matching exercise on a data set that has good temporal coherence/

So yeah, hefty initial calculation with lower latency interim calculations. Moving to another room, kicking a table or a wall falling down is going to cause a glitch but you'd probably be thinking about other stuff if that happened

There's lots of opportunity here. I really hope the onboard special processing unit they have is programmable on a very low level though rather than a static but of HW.

[gravy]

It’s what they call “pinning” that I’m skeptical of. Where you move around and it captures that information, does all the calculations, renders it and outputs it to the device so quickly that it keeps up perfectly. They can do the calculations very quickly but there the limitations of the input device and the renderer are hard to eliminate.

The pinning could work in a stateful way if you assume the environment you're projecting onto isn't going to change much (ie tables and walls don't move)

So an initial scan could have higher latency but after that orientation ends up being a much simpler pattern matching exercise on a data set that has good temporal coherence/

So yeah, hefty initial calculation with lower latency interim calculations. Moving to another room, kicking a table or a wall falling down is going to cause a glitch but you'd probably be thinking about other stuff if that happened

There's lots of opportunity here. I really hope the onboard special processing unit they have is programmable on a very low level though rather than a static but of HW.

[Little stoned kiwi]

Augmented reality is more interesting to me than virtual reality. It has the potential to be an immensely disruptive, world changing technology.

[gravy]

Yeah, and it's a superset of VR really. I'd still really want a black out mode on something like this as well AR. Hope that's a feature the put in this

[benedictw]

Heres a question - how are they doing the light projection - is this a added light (lasers, LEDs) solution or is there some support for darkening? If its added light then it can't let the scene get darker and thats really going to fudge the integration side of things. And all your AR things will be moderately transparent...



That has the buildings etc being slightly transparent and the white door at the top comes through fairly clearly, so even though its a bullshot I think we're looking at light adding rather than appearing solid. (Think leia hologram minus the flicker). (Might explain why the headset is very 'ski goggles dark tinted').

Aside from the usual AR gee wizz, the most useful use case is just giving me a curved 8k display that wraps round my workspace.

They've got their announcement out before Magic Leap so I'm going to say that multiple companies think this sort of tech is going to be rather useful.

[gravy]

Projective AR issues laid out pretty clearly hear by Abrash from when he was at Valve

http://blogs.valvesoftware.com/abrash/w ... time-soon/

And yeah, additive only is likely what they've chosen and an issue too