Hololens tracking isn't nearly good enough, in my opinion, and isn't going to get there soon. Vision algorithms + and inertial tracker could be interesting though. Appropriate tech exists, but is currently expensive.
There has been good work on eye tracking in other spaces for many years now, so that seems a much cleaner implementation path to work well.
edit: I'm going to add a caveat, I mostly care about VR and AR in decidedly non-consumer applications, so my bar may be unreasonably high for many consumer applications.
What did you find wrong with the Hololens tracking? I use it a couple times a week and have been super impressed with it so far. Seems to very solidly lock on locations, and I can walk to the far end of the office and my content is still right where I left it. You can use is in huge unbound areas and it doesn't mind at all.
The Hololens still has a lot of issues, the FOV's crappy, hand tracking needs a lot of work, occlusion isn't great, etc. but tracking is one place that they're top of the game.
It's probably fine if you just want "if I go back to my desk, this window is on the left hand side". If you want more precision and accuracy, or consistency with physical measurements/other tracking systems, it's not so good in my experience. Happy to revise/update that experience though!
That kind of precision isn't necessary for tracking position accurately for VR/AR though, Hololens' current system is pretty damn good and I'm sure internally it's improved a bunch as they roll out their own VR headsets too.
It absolutely is needed for many AR applications. VR is a bit more forgiving, particularly if you are talking consumer and games. Any time you need maintain a relationship to the physical space you are in you can run into trouble though.
Re:VR
by RangerScience on Friday January 13, 2017 @02:05PM
(#13393672)
Which eye-tracking tech did you try at SIGGRAPH?
The SMI one, and the accompanying Nvidia foveated rendering demo.
Re:VR
by RangerScience on Friday January 13, 2017 @02:59PM
(#13394199)
Were they in VR headsets (or glasses), or were they the desktop bars?
(Last time I tried a bunch of eye tracking, the desk-mounted was amazing and the head-mounted wasn't)
They were inside the headset, lining the lens, such that you couldn't tell between a normal Vive and an eye-tracking Vive unless you cracked it open.
- Eye tracking (it already works perfectly, I've tried it myself at SIGGRAPH), this will enable a few cool things:
- Inside-Out Tracking - using computer vision to provide 6DoF tracking for headsets without the need for external trackers. Will allow mobile headsets to have positional tracking (which is SO VERY important for VR) and will allow desktop headsets to have lower setup complexity (less important). - note: Microsoft will likely dominate this by my guess, seeing as probably the strongest part of the Hololens is it's excellent tracking.- Wireless adapters for existing headsets - these made a big splash at CES and apparently work pretty well. Making the existing experience un-tethered will definitely help room-scale experiences.
- Self-Contained headsets - this is vital to mass-adoption of VR imo. I think we'll see some of these this year, though probably not from HTC/Oculus yet.
AR, while definitely more the "consumer" product in the long run, is still far off as the display tech just isn't there yet. But the above advancements in VR pave a way for AR in the future, until there is no longer a distinction between them device-wise, but it rather becomes a slider of "how much reality do you want to replace?".