Why not put the camera on the oculus itself?

"raidho36" wrote:
Well, dedicating to seated position gaming experiences is reasonable. Standing upright while wearing the Rift imposes some serious amount of danger, since you're instantly losing your entire spatial awareness reserve. You will have zero clue where is forward, where is north, where is your furniture and where are people around you. This alone can get you some nasty bumps against "invisible" obstacles that can potentially lead to wrecking the obstacle, the Rift, or both, and can cause health damage, but combined with VR experiences it yields even more such danger. You don't want to break your Rift (and your face) because you tripped over a chair, or because you slammed your head into the wall (or housemate's face) full-tight while trying to dodge virtual fireball.

How about this.
It will require some skills but possible for room /garage size environment:
a) Take apart Gametrak .
b) Solder extension wires to separated string spool modules.
c) Clamp modules at top edges of safe area in a room with strings connected on user (top of the head perhaps to avoid entanglement).
d) Write small app to give sound alarm when user reaches edges of safe area based on data from game track.
or make something more complicated to give opaque "safe area" borders in vr environment.

The example below a bit abstract ( sound=data) but I hope it will give idea what I'm trying to say.

I'm with Adaptation on this one...

Saw this coming too... take computer vision and stereo camera's, make it fast and low latency.

If so many things happening point in this direction... there's kinect fusion, building a near real time 3D dataset. Apple buying primesense, and now there's Google project Tango: https://www.google.com/atap/projecttango/

Please Oculus take note !

In the same way how the rift became possible with the availabilty of cheap good panels from mobile phone developement, positional tracking will follow the same path.

In the same way as tracking had too much latency before, then Oculus codeveloped this super fast dedicated chip to do it, they could develop a dedicated chip for this computer vision positioning.

Actually that is exactly what project tango uses:

http://techcrunch.com/2014/02/20/inside-the-revolutionary-3d-vision-chip-at-the-heart-of-googles-project-tango-phone/

This way solves several purposes at once.

You get positional tracking. (i suppose that a device that does realtime 6D slam, inherently knows it's own position and orientation inside that 3D dataset.)

You get the outside world in the headset. in more interesting ways possible then with just a video feed, since you have the whole 3D representation.

both will be needed for AR systems anyway...

for developers it creates new and interesting possibilities in game play. users can 3D scan a real object and use it inside the virtual world immediately. Maybe if this aproach advances enough it can be used as input as well.

There could be some dedicated memory in the headset, so you calibrate the room you use it in most often, by looking around, and that data is saved for future use. faster units in the future will do this instantly and for bigger area's, or you just don't keep all data around, only what is needed for postioning and not bumping into things.

One dificult thing with having stereo video camera's on the front could be aligning the camera's and optics and such to have a 1-1scale of the outside in the rift. but optics and warping is exactly what set palmer and oculus apart to create the rift to begin with! It's essentialy the same method but in reverse.

Going further, if the stereo camera's capture, create a 3D dataset in realtime, even this alignment is irrelevant, since you can render and display the reality out there from that data. mixed in with the virtual world or not. Sounds creepy i know... like reality virtual.

only things keeping this aproach back is the speed and latency in which this could work, but then, that is just a matter of time, just like it is/was with the high res screens...


i guess this would be a big stretch for CV1 but i hope they investigate this !


Oh, and then put in a tegra K1 and drop the cord...

PS leave sony play it's own game with the PS4 and whatever they come up with, Oculus can create it's own dedicated platform, which seems they're already doing... Not that i have something against sony, it would just be a pity to see them copying or running you over. I hope there will be some cooperation or just healty competition and people get choice.

"FrederickD" wrote:

Saw this coming too... take computer vision and stereo camera's, make it fast and low latency.

You're talking about SLAM (Simultaneous localisation and mapping), which when done by a IR field as in kinect (and I believe Project Tango) is much easier and more reliable.

"FrederickD" wrote:

Please Oculus take note !

I guarantee they are already hard at work on this, and I believe it is the future of wireless/camera-less tracking in a self contained headset unit. The technology and processing power necessary is in its infancy though. Project Tango is a good start.

"FrederickD" wrote:

You get the outside world in the headset. in more interesting ways possible then with just a video feed, since you have the whole 3D representation.

both will be needed for AR systems anyway...

I like your train of thought, and I agree that is the future, but it's going to take some time. Definitely not CV1 stuff, but probably not too far in the future.

There must be a way to use more than one camera? I can imagine something like "The Virtualizer" with a camera on everyone of the three rigs supporting the rifter using it. Or perhaps you could use a camera in the roof? Would that be able to track you 360 degrees if you are on an omni? There are also cameras that can film 360 degrees, but whether these can be used for motion tracking or if they are cost effective I don't know. Perhaps it can be done with sonar? Just brainstorming here.

(This is a link to the Virtualizer, just in case you haven't seen it yet. I don't think it's as popular as the Omni. - http://www.youtube.com/watch?v=qyzOzqofUME)

Virtualizer, Omni mehh them's just childrens toys.

I want this hooked up to an oclulus.

https://www.youtube.com/watch?v=bxbjZiKAZP4

"przecinek" wrote:
Someone should tell Oculus that sitting experiences are just not immersive enough ... how can you seriously consider VR for sitting purposes? Just fire up any FPS with Oculus on and see the difference between when you sit and stand. In the first case you are just playing a full 3d game, whereas when standing you trick your brain in believing you are actually there!

I guess a lot of us will have to use wireless hydras to get a full 360 degree pos tracking.

usb cameras are 5$ for good ones from ebay... lifehacker.com/5850388/use-a-cheap-ikea-lamp-to-build-an-affordable-desktop-camera-jig can lower it to sit and spin or raise them when using a treadmill.. need software to know that a dot not in vision of one camera will be in vision of one of the other cameras..

having 2+ micro/mini usb slots on the DK2 camera will allow it to be in the middle and reduce delay from a standard daisy chain where the end camera is slowest to input (however slight). 3 wide angle cameras should cover a single person with a treadmill and some cheap light stands

That demo video on project tango looks absolutely amazing, and would coincide perfectly with the game idea that I've been playing around with.

With something like this mounted to the face or the top of the rift, you could track your positional data in real time, and hopefully there will be a way to solve the latency problems between the devices.

Heck, just knowing that this technology is out there is amazing. If I were even close to the level of john carmack, i would be taking that hardware out of the phone entirely and integrating it with the rift. The possibilities that this unlocks for the world of AR is amazing.

If you just had simple stereo-vision cameras hooked up to the front of the rift, along with this scanning device, you could use the scanning data for collision and supply the visual data through the simple cameras, adjusted for IPD etc. of course.

This would completely cut a lot of rendering out of the equation, and tons of the problems we have with rendering an entire virtual world would pretty much evaporate for AR purposes, at least the way I'm envisioning it. This way, you would only have to render the changes you are making to the AR version of the world, not necessarily the entire scene in full.

There would be a lot of work involved on tracking spots where changes are being applied to the AR world, and syncing these between devices, and the rest of the world.

In short, A lot of AR type things will be possible in a much shorter time than I thought.

camera on the rift is great for augmenting reality.. but to use it for positional tracking **seems** to mean you get to repaint everything around you in a zebra print QR code..

having the DK2 include a central camera could mean you could add USB cameras as required to increase the turning radius of your wireless modified rift.

having the single central camera that comes with the rift work as the hub for the wireless rift would be optimal..

http://www.roadtovr.com/oculus-rift-camera-mod-lets-you-bring-the-outside-world-in/ would be fun so long as that's what the affixed camera is for.. maybe add picture in picture when the optical walls are broken instead of seeing a floating 1280x1080 screen surrounded by empty

*edit*added "seems" to the part about seems to mean.

"adaptation" wrote:
Virtualizer, Omni mehh them's just childrens toys.

I want this hooked up to an oclulus.

https://www.youtube.com/watch?v=bxbjZiKAZP4

I tend to stop development well before the point where a bug can kill a person outright.

"jherico" wrote:

"adaptation" wrote:
Virtualizer, Omni mehh them's just childrens toys.
I want this hooked up to an oclulus.
https://www.youtube.com/watch?v=bxbjZiKAZP4

I tend to stop development well before the point where a bug can kill a person outright.

Obviously the motion engine would need to be abstracted from the application layer. A software bug in a car can kill you just as dead as a robot, soon cars will be driving us on there own. Building human interaction for something like this in a safe was is important and it's not impossible. High end folks have been doing it for years too, I picked the other video because safe is intrinsically more boring than anticipating disaster.

BTW these robots (even in the first video) all have mechanical breaks that can be activated automatically along with cutting power to the motors if the robot leaves a defined operating envelope. Done properly they are mufti fault tolerant and can me extremely safe.

https://www.youtube.com/watch?v=QCGcWOwV8Qs