cancel
Showing results for 
Search instead for 
Did you mean: 

Are other then IR-Camera pos. tracking options considered?

ElectricMucus
Explorer
For starters, I don't really like DK2's tracking. It's accurate only within a small area and the mounting of the camera is very restrictive. For instance if you want to use the Rift in the same position as you were sitting at your regular desktop you get something I'd call the "Monitor hole": Since you have to position the camera facing your front side and most likely you won't put it in front of the monitor or even on top of it you get an area where you don't have any tracking.
It would be somewhat ok if the camera would just work at the same distance the average monitor is, but it's FOV is too small for that to result in a good tracking volume.


Now, I know the CV1 camera might be better, have a better FOV or better resolution, frame rate, and have Ir LEDs on the back. But the principal issue would remain, namely the requirement environment where line of sight for a fixed object has to be maintained..
Have I been missing something, but why aren't oculus looking into magnetic tracking, radio frequency tracking (RFID, WIFI, GPS, etc..) or similar stuff. I also would like to point out one thing: I think People don't have a problem with no inside out tracking or a fixed tracking volume, it's the obstructed tracking what bugs us.

Playing Devils advocate here..
Acquire or talk with Sixense (Stem)?
8 REPLIES 8

Fredz
Explorer
"ElectricMucus" wrote:
It's accurate only within a small area

How did you get to this conclusion ?

"ElectricMucus" wrote:
Have I been missing something, but why aren't oculus looking into magnetic tracking, radio frequency tracking (RFID, WIFI, GPS, etc..) or similar stuff.

They have. Hybrid optical/inertial is the best technology for accurate/precise and low latency tracking. Name one technology that can deliver 0.05 mm precision with ~1 ms latency.

Even high end magnetic tracking systems costing several thousands of dollars (Polhemus, Ascension) can only deliver an accuracy/precision of 0.5 mm (10x worse) and ~4 ms latency (4x worse).

"ElectricMucus" wrote:
Acquire or talk with Sixense ?

It's not available yet, while the DK2 has been available for 6 months. There is no reason to think that it'll be more accurate/precise and with similar latency than the DK2 tracking system.

Also it's $300, almost the price of the whole DK2 (HMD + camera). Bundled with it the DK2 would cost $600. That doesn't make sense.

cybereality
Grand Champion
Just like to add that Oculus tried a whole bunch of crazy stuff before settling on the optical tracking solution.

There is a reason most games and movies use optical for motion-capture.
AMD Ryzen 7 1800X | MSI X370 Titanium | G.Skill 16GB DDR4 3200 | EVGA SuperNOVA 1000 | Corsair Hydro H110i Gigabyte RX Vega 64 x2 | Samsung 960 Evo M.2 500GB | Seagate FireCuda SSHD 2TB | Phanteks ENTHOO EVOLV

ElectricMucus
Explorer
Freds,

Based on my personal use with the DK2 I had issues with the camera and it's so called precision if the camera is more than two meters away. Unity apps even stop tracking after that and display a red message, with a reason, it's quite discomforting.
I think that the practical precision of the IR camera tracking is way lower than the advertised 0.5mm.
Magnetic Tracking aka Sixense doesn't cost much, in parts at least.. just in IP and that can be negotiated.

PS: Don't split quote in forums, you don't look cool if you do since this isn't a mailing list and it's difficult to reply back in context.

cybereality,

I've read that you've tried many different things. My suspicion is that you guys suffer a little under the Not Invented Here syndrome. 😉

But yeah, motion capture with optical systems, let me throw that right back at you: They use several cameras from multiple angles simultaneously, so how about it: Provide versions with two and three cameras with the CV, adopt the SDK to utilize multiple cameras and provide means to purchase extra cameras for existing DK2 developers and I'm fine.
For the case of the Rift there are significant diminishing returns though. So that's why I suggested looking more at other possibilities.
The other reason is home environments are way more crowded then motion capture chambers...

Fredz
Explorer
"ElectricMucus" wrote:
Freds,

It's Fredz, not Freds.

"ElectricMucus" wrote:
Based on my personal use with the DK2 I had issues with the camera and it's so called precision if the camera is more than two meters away. Unity apps even stop tracking after that and display a red message, with a reason, it's quite discomforting.

That's what it should do, the camera has a 0.5m - 2.5m range. You're most probably past the maximum range.

"ElectricMucus" wrote:
I think that the practical precision of the IR camera tracking is way lower than the advertised 0.5mm.

It's 0.05 mm, not 0.5 mm. You "think" that the precision is lower but you didn't measure it, did you ? Also this kind of precision doesn't come only from the camera, but also from the inertial sensors. The camera is used to avoid the drift from the sensors. That's why I talked about hybrid optical/inertial and not optical only.

"ElectricMucus" wrote:
Magnetic Tracking aka Sixense doesn't cost much, in parts at least.. just in IP and that can be negotiated.

Can you back this up with a reliable source or is it only a speculation ?

"ElectricMucus" wrote:
PS: Don't split quote in forums, you don't look cool if you do since this isn't a mailing list and it's difficult to reply back in context.

I've been doing this for the past 20 years because it's more readable and allows for a more articulate discussion. I won't change for you. Also, I couldn't care less about looking "cool".

"ElectricMucus" wrote:
I've read that you've tried many different things. My suspicion is that you guys suffer a little under the Not Invented Here syndrome. 😉

Oculus VR just acquired Nimble Sense, what did you say about the NIH syndrome ?

JustinL
Honored Guest
Optical tracking is clearly the best option for the precision and speed reasons listed above. If occlusion and capture volume are a problem, the solution is pretty simple: add more cameras and position them properly to maximize the volume. GPS, magnetic, wifi, etc. have bigger problems that are harder to overcome than optical. This is a case where the simplest solution is definitely the best.

VizionVR
Rising Star
Well, now that Oculus is officially working with 13th Lab, it's only a matter of time before an inside out tracking system is developed. This will lead to infinite tracking that's even possible with the Samsung Gear.
Not a Rift fanboi. Not a Vive fanboi. I'm a VR fanboi. Get it straight.

MrMonkeybat
Explorer
magnetic tracking is not entirely occlusion free, metal objects and dense objects can interfere and requires a bit of calibration and drift correction. What something like Stem could probably do with are some leds for calibrating and zeroing with the DK2 camera.

Cresent Bay demonstrated a much larger tracking volume from a camera which was both wider angle and longer range. With leds on the back I dont think anyone trying it manged to loose tracking.

Oculus just acquired Nimble which where going to release a depth camera, so that might might give them more accuracy in the depth dimension also, and perhaps inside out tracking when combined with 13thLab's stuff.

ElectricMucus
Explorer
Fredz, you're an annoying drone.

"mrmonkeybat" wrote:
magnetic tracking is not entirely occlusion free, metal objects and dense objects can interfere and requires a bit of calibration and drift correction. What something like Stem could probably do with are some leds for calibrating and zeroing with the DK2 camera.

Cresent Bay demonstrated a much larger tracking volume from a camera which was both wider angle and longer range. With leds on the back I dont think anyone trying it manged to loose tracking.

Oculus just acquired Nimble which where going to release a depth camera, so that might might give them more accuracy in the depth dimension also, and perhaps inside out tracking when combined with 13thLab's stuff.


That's good to hear, a depth camera would at least work independent of rotation. With the DK2 as long as you are looking straight ahead there are plenty of LEDs visible and the tracking works fine. Now if there is a way to actually measure the distance to the users head, it won't "guess" the wrong position after consecutive rotations which also have translations.
The occlusion issue however won't be solved unless there are at least two cameras.