360 positional tracking with dk2

The other thing I tried was attaching the camera to an arm extending from my chair. This way the camera was always a couple feet out directly in front of the center of my body (even when turning). This theory worked great, but when you turn the chair the camera tries to figure out where it is and jumps.

The camera can't tell if you are turning your head or if the camera is moving. I think I'd need to calculate the rotation of the chair separately in order to input my own camera location, rather then using the camera location from the sdk. The only way I can think of doing this would be through an accelerometer attached to the chair. Would calculating this through a sensor and arduino then communicating to the game cause too much latency? Seems like it should be faster then the camera can pull and calculate each frame.

Any ideas how I could bypass the accelerometer altogether? Maybe something where I set an initial camera position and then it is calculated during the rest of the game play by the camera and dk2 sensor data?

Wishful thinking... but would be cool of Oculus just allowed you to buy additional cameras and set them around the room :D

You don't need a rear camera.

All you need are markers at the rear of your head, so when you've turned your head 180°then those rear markers are picked up by the camera.

Yeah, I personally think that come the consumer version we'll see two specific updates.

One, will be an extension of the IR tracking to the the rear of the device.

Due to the shape of the Rift (sticking out from your face) tracking is pretty good even slightly past 90 degrees so even just two IR leds in the back middle of the device could be feasible.

I know Oculus have said they're focusing on the seated experience but the truth is it's the first that that breaks for most if you've got the camera set up well.

Following on for that I think the second thing I see being improved is the camera field of view.

While completely adequate for development, for a consumer product I really think it needs to have a wider FOV to fit easier into peoples current setups.

I think most of us have had to make specific changes to our setups in terms of monitor position (or tripods) to make the camera fit that (without understanding the technical details as to whether it could or couldn't work) I think would make it a lot easier to use.

Sony Morpheus got positional tracker on the rear side :



I think it will be the same with the CV1.

Have you tried mounting camera directly above you, looking down at 90degrees? It should be able to track 360 to some degree.

Would be keen to see a result of that test! Means we can use with omnidirectional treadmills.

I've tried with the downward camera view, but the tracking was really glitchy. I think it was looking for the front of the HMD and so it didn't understand what it was seeing. It's worth testing some more though.

Sony Morpheus got positional tracker on the rear side

Sure that arend just fancy Lights like in the Front of Morpheus. ?

Those 'fancy lights' are the Morpheus' tracking markers: it tracks using the PlayStation Eye, which not being an nIR camera uses coloured markers instead (same as the PS Move system).

With the ski-goggle mount, you can't easily 'just add' a rear marker constellation. The Morpheus has the rear section rigidly attached the the front section, so the whole thing can be tracked as a rigid object. Without that rigid attachment, you not only need to look for markers and use them to infer the marker position, you would need to add some additional tracking method to know the relation between the rear marker and the optics (i.e. the part you;re actually interested in the position of).

Man, the power of the tracking on the DK2 has turned out to be the biggest surprise for me..
I really couldn't be any more impressed.

I did find the DK2 manual slightly misleading about proper camera setup-

IMAG0208_1.jpg

It's funny because if you take the above image at face-value, this guy is aprox. 15ft. tall !!

So when you factor in real-world scales, mounting the camera directly in front of you on the monitor isn't really an effective way to develop- unless you want to roll back 4 feet every time you check content.

For me, what I found works well is to actually have the camera off to the right or left of you at a 45-55 degree angle.. I also have it pointing down at you somewhat from above at maybe 20 degrees.. In that configuration it tracks 360 degress relatively well, and you also have an incredibly wide area to work with. The 45 degree angle gets rid of the blocking effect of the monitor, and I figure most people don't have a lot of room behind their monitors anyways.

To play around, I've been using a mono-pod with a clamp to the back of a chair.. It's a cheap way to play with all angles.

Also the 'camera bounds' feature within the desk scene is incredibly useful for set-up... Nothing seems to be more important than having the oculus contained in the rear portion of the bounded camera area. Then if configured well, it goes from simple head-tracking, to really allowing you to stand up or even get down on the ground.. Very much like the ValveVR rig.

two thumbs up oculus!!!
Love it-