Why not put the camera on the oculus itself?

At the start, very few people will want to do anything but sitting VR. It's not something you can really do without something like the Omni, which is way more involved than most people want to get.

Oculus is optimizing for the 99% who will want to do it sitting down, not the 1% who want to dedicate a room of their house to VR...

I think the key points making this idea less plausible are price - keeping an affordable retail product,
time/lag - the camera feed would have to be processed on the pc may introduce a difference in lag between rotational movement and positional tracking
complexity - so you have a dot on a wall. How does this become a known position? Is the Ir blaster measuring the distance in real time which may again be "slower" in comparison. Or scan beforehand? what happens if your own body blocks the beams?

I'm not saying it's not an idea that couldn't work. Just that OVR probably tried it or something similar and decided it was less feasible than the static camera.

The proposed CC solution is cheaper, less lag? Less complex as the Ir dots are in a known grid and never deviate from their relative position regardless of the room they're placed in.

There is absolutely nothing stopping people standing up to use a crystal Cove solution. If the camera fov is limiting you raise the camera? You just can't walk off around the room, maybe you could take a small step but hey your tethered anyway.

Just my opinion :)

In another thread someone used live webcam feeds mounted on the oculus and the delay was rather unusable. You could couple the hardware directly from the camera to the display and get the latency down to almost nothing. Then
using stereoscopic computer vision aided with the orientational data from the rift you could easel make a 3d model of the environment to selectively overlay your augmented reality content onto the live video feed.

Entirely doable but it would add substantial development time another $100+ in parts because you need high quality cameras with fish eye lenses, and probably an fpga based video distortion engine on the device.

Alternatively you could take a simple usb camera approach and feed it into the computer vision to generate the 3d model and feed the 3d models data into the rift, this is the method being developed for high latency teleoperations of equipment and robots. There was a cute game based on the concept where you where a robot that had to navigate a maze as fast as you could but there was a large delay programed into new information so you would move ahead like a ghost of the robot but your vision would be deprived of new information until the real robot advanced.

Anyhoo this cheap and dirty way of doing it could be supported for ~$20 of hardware with almost no hardware development time required. Low cost cameras could be used. Initially it could be used to augment the orientation data with positioning information, as standing up and sitting down or leaning is slower than whipping your head and carries more inertia (more predictible) the update rate could be something even cheap hardware could accomplish. Actually using it for augmented reality type setting could be added later in software so even software development delays would be minimal.

Sorry for nudging an old thread, but this was the closest match while searching for someone who may have already explored this idea:

I'd considered the possibility of a greenscreen-style fabric, but instead of using it as a background, invert the idea. Have a greenscreen colored long-sleeve pullover (shirt) combined with a pair of down-angled cameras on the Rift as a very low-tech means of putting the players body into the scene.

This would allow you to look down and see a size-accurate, position-accurate representation of your arms and upper torso in-game. As it is, it's rather odd to look down and find yourself completely disembodied. Sure, you can add a player model into the game for you to see when you look down, but without some kind of detailed body tracking, the limb/body proportions and positioning will be off. Tracking hand position and orientation w/ something like the Hyrda isn't sufficient either, as you can only guess how the limbs are oriented relative to the device.

Using a shirt with a trackable color combined with cheap/low-res stereo cameras just seemed like a viable means of address this. Am I totally off my rocker or is there something worth exploring there?

FWIW, I'm just a long-time coder, but gaming software n00b waiting for my dev kit to arrive, and this is the sort of thing that's popping into my head :) Can't wait to dig in!

This is an interesting idea.

Packages like Boujou and SynthEyes calculate camera movement by dispersing dozens of feature trackers over the video.



The process is fast but perhaps not fast enough for real-time at 60fps. But if it was, you could mount a camera like the XBox Kinect (which can shoot both IR and RGB) to the Oculus, which feeds into a feature-based camera tracker. You wouldn't need to position markers all around your room this way.

:9

(EDIT: Come to think of it; If you look around the forum, you'll find a thread about an optical hand capture solution, that uses a colour-coded glove.)

Why can't you put the markers on that massive computer monitor that you have behind your keyboard?

One in each corner should do it?

I know! How about you paint markers on every member of your family and have them standing around at random points in your basement! Lets keep this thread goingggggggggggggggggggggggggggggg!!!

Problem solved! Why aren't you working for Oculus?!??

"adaptation" wrote:
In another thread someone used live webcam feeds mounted on the oculus and the delay was rather unusable. You could couple the hardware directly from the camera to the display and get the latency down to almost nothing. Then
using stereoscopic computer vision aided with the orientational data from the rift you could easel make a 3d model of the environment to selectively overlay your augmented reality content onto the live video feed.

Entirely doable but it would add substantial development time another $100+ in parts because you need high quality cameras with fish eye lenses, and probably an fpga based video distortion engine on the device.

Alternatively you could take a simple usb camera approach and feed it into the computer vision to generate the 3d model and feed the 3d models data into the rift, this is the method being developed for high latency teleoperations of equipment and robots. There was a cute game based on the concept where you where a robot that had to navigate a maze as fast as you could but there was a large delay programed into new information so you would move ahead like a ghost of the robot but your vision would be deprived of new information until the real robot advanced.

Anyhoo this cheap and dirty way of doing it could be supported for ~$20 of hardware with almost no hardware development time required. Low cost cameras could be used. Initially it could be used to augment the orientation data with positioning information, as standing up and sitting down or leaning is slower than whipping your head and carries more inertia (more predictible) the update rate could be something even cheap hardware could accomplish. Actually using it for augmented reality type setting could be added later in software so even software development delays would be minimal.

One solution we can use, is the one the Google phone will use.

You have a camera to track the movements
You have a laser to mesure distance of what you're camera see.

With those both information you can recognize your movement.