Meant to be Seen

bobjwatts wrote:@2EyeGuy - Does it do the split blue/red style 3d that is in the youtube video, on top of the standard side-by-side.

What on Earth are you talking about?
Cybereality's Vireio Perception driver might support blue/red style 3d glasses for people who don't have an Oculus Rift.
But if you are talking about the Oculus Rift (Virtual Reality Head Mounted Display), then you aren't making any sense, and I think you completely misunderstand something.

Ok.

The Anaglyph 3D in the video is for non-rift devices, got it. Thanks 2EyeGuy.

cybereality wrote:Also, just note that FRAPS for some reason introduced serious lag/delay in the head-tracking so this made it not as smooth as normal.

Why not use something like "recorddemo DEMONAME, stoprecording, playdemo DEMONAME, avidemo DEMONAME" in Doom3, then convert it to a video? No extra lag by any software, video convertion or hdd IO trying to record a video at the same time the computer is rendering it.

A naive question - is it possible to somehow leverage 3d vision here. For my 3dtv i have an inf which tricks 3dvision into outputting checkerboard. As far as I understand checkerboard it should be possible to convert it to SBS using shaders, apply warping and mirror that to oculus rift. This would effectively give almost native 3d support. The fov and head tracking will still need to be handled separately though

Quant wrote:A naive question - is it possible to somehow leverage 3d vision here. For my 3dtv i have an inf which tricks 3dvision into outputting checkerboard. As far as I understand checkerboard it should be possible to convert it to SBS using shaders, apply warping and mirror that to oculus rift. This would effectively give almost native 3d support. The fov and head tracking will still need to be handled separately though

3dvision supports checkerboard because of the launch Mitsubishi TV's, which needed a way to claim to support 1080p 3d on the old HDMI standard. So the inf you got probably just says use the Mitsubishi profile Have no clue if this could be converted to side-by-side fast enough to be used . . .

2EyeGuy wrote:

bobjwatts wrote:@2EyeGuy - Does it do the split blue/red style 3d that is in the youtube video, on top of the standard side-by-side.

What on Earth are you talking about?
Cybereality's Vireio Perception driver might support blue/red style 3d glasses for people who don't have an Oculus Rift.
But if you are talking about the Oculus Rift (Virtual Reality Head Mounted Display), then you aren't making any sense, and I think you completely misunderstand something.

Ok I really get it now, took me a while. After watching Cyber's older youtube video so many times side-by-side I - through some sort of brain meltdown- recalled the new video side by side and anaglyph, after more research into the subject and then watching the new video again I now see the video was not side-by-side and I am crazy.

I apologise for wasting space in this thread, although it did prompt me to learn some more, including how cool the youtube 3D controls are.

Please, don't mind me, carry on...

Hello.
I'm joining the party. I'm very excited to see good progress on your driver Cybereality.
I've watched the 3D video for regular 3D displays. It looks quite good. It's not perfect of course since it's still heavily in development, I am confident you'll overcome most issues.
As others have mentioned the delay between the left and right eye is a problem, but I don't agree with the "fast vs slow movement thing" It a problem in ANY situation, and should never be downplayed : especially at low framerates like shown in the video (Fraps 30fps caps). I have had this issue with my dual-projectors system and I cannot stand to have a delay between the eyes. And It would be made worse by the games I play, most of which don't reach 60fps in 3D.
Supporting advanced stuff like head tracking and custom aspect ratios is important for the Rift. And getting the shaders to render properly for each game is also a very important step. But in my opinion, I think you should focus more on getting the basics right in the first place by making simultaneous rendering a higher priority on your list.

The 6DOF head tracking thing is nice, but it looks really weird in the video, whenever you turn your head it feels as if everything is heavily distorted. Perhaps because of the type of 3D projection used which naturally distorts stuff at high FOVs. We are used to seeing stuff distorted one way when the camera is horizontal, and then the distortion becomes completely different when you tilt your head. There's also the issue of doing actual head tracking on a regular 3D display rather than an HMD.
On a regular 3D display, the display doesn't follow your eyes, so when you tilt your head you'll also have to tilt the interocular vector to match the position of the viewer. Finally : some 3D displays don't support head tilt, so you might need to provide an option to lock head tilt (provide only 5DOF instead of 6)

BlackShark wrote:The 6DOF head tracking thing is nice, but it looks really weird in the video, whenever you turn your head it feels as if everything is heavily distorted. Perhaps because of the type of 3D projection used which naturally distorts stuff at high FOVs. We are used to seeing stuff distorted one way when the camera is horizontal, and then the distortion becomes completely different when you tilt your head. There's also the issue of doing actual head tracking on a regular 3D display rather than an HMD.
On a regular 3D display, the display doesn't follow your eyes, so when you tilt your head you'll also have to tilt the interocular vector to match the position of the viewer. Finally : some 3D displays don't support head tilt, so you might need to provide an option to lock head tilt (provide only 5DOF instead of 6)

Not sure if BlackShark is describing this correctly, but he brings up a good point. Cyber, have you given some thought to head and neck models? I assume right now for simplicity you are using a simple camera rotate. But the center of rotation dramatically effects the look of the transform. I used one of Carmack's earlier head models where it seems like he modeled rotation around the nose. So when I rolled my head around the chin (flopped my head onto my shoulder) it creating an eerie shearing effect on the world that broke the illusion of reality - or at least made it look like I was on drugs. It was really striking how different the world looked when I switched between nose and chin rotations. Anyway, just a head's up to think about this. It's possible the Oculus SDK may even include a configurable head and neck model for their tracker values which would be convenient to use and would provide some consistency to many Rift games.

rowanunderwood wrote:

Quant wrote:A naive question - is it possible to somehow leverage 3d vision here. For my 3dtv i have an inf which tricks 3dvision into outputting checkerboard. As far as I understand checkerboard it should be possible to convert it to SBS using shaders, apply warping and mirror that to oculus rift. This would effectively give almost native 3d support. The fov and head tracking will still need to be handled separately though

3dvision supports checkerboard because of the launch Mitsubishi TV's, which needed a way to claim to support 1080p 3d on the old HDMI standard. So the inf you got probably just says use the Mitsubishi profile Have no clue if this could be converted to side-by-side fast enough to be used . . .

Yeah, it's just tricking into generic dlp. As far as I understand the checkerboard is just LRLR pattern so converting to SBS is just a question about separating odd and even pixels. Or maybe i'm missing something

brantlew wrote:

BlackShark wrote:The 6DOF head tracking thing is nice, but it looks really weird in the video, whenever you turn your head it feels as if everything is heavily distorted. Perhaps because of the type of 3D projection used which naturally distorts stuff at high FOVs. We are used to seeing stuff distorted one way when the camera is horizontal, and then the distortion becomes completely different when you tilt your head. There's also the issue of doing actual head tracking on a regular 3D display rather than an HMD.
On a regular 3D display, the display doesn't follow your eyes, so when you tilt your head you'll also have to tilt the interocular vector to match the position of the viewer. Finally : some 3D displays don't support head tilt, so you might need to provide an option to lock head tilt (provide only 5DOF instead of 6)

Not sure if BlackShark is describing this correctly, but he brings up a good point. Cyber, have you given some thought to head and neck models? I assume right now for simplicity you are using a simple camera rotate. But the center of rotation dramatically effects the look of the transform. I used one of Carmack's earlier head models where it seems like he modeled rotation around the nose. So when I rolled my head around the chin (flopped my head onto my shoulder) it creating an eerie shearing effect on the world that broke the illusion of reality - or at least made it look like I was on drugs. It was really striking how different the world looked when I switched between nose and chin rotations. Anyway, just a head's up to think about this. It's possible the Oculus SDK may even include a configurable head and neck model for their tracker values which would be convenient to use and would provide some consistency to many Rift games.

As a future enhancement, I wonder if it would be possible and/or practical to design a 'jacket' that has a Hillcrest or equivalent sensor at the mid-point of all upper body joints (one or two on your spine, in-between your hips and neck, one as close to the center of your eyes as possible, one at each shoulder rotator cuff, mid upper arm, mid forearm, etc) and to use the combined data from those sensors to figure out exactly where all your body parts are with respect to each other using basic trig. I would imagine that some level of calibration would need to be done that involves taking physical measurements of all your limbs (for the purpose of determining where the sensors are WRT each other), and then mapping those values into the game's measurement system.

This setup would be for games where you're seated, but the concept could easily be extended to standing games by adding more sensors below the waist. Doing so would provide full body position determination WRT a fixed point (namely, your feet). The only movement that would require other means to detect would be jumping (no sensor angle would change during a jump)

This seems like such an obvious solution that it has to have been considered and discarded for some reason.

Thoughts?

Mel wrote:This seems like such an obvious solution that it has to have been considered and discarded for some reason. Thoughts?

Answer : cost ($4000 for full body tracking).
See : http://www.mtbs3d.com/phpBB/viewtopic.php?f=120&t=15564
Would be less expensive for just the upper body, but still nowhere near a consumer budget.

EDIT : sorry RoadKillGrill...

Mel wrote:As a future enhancement, I wonder if it would be possible and/or practical to design a 'jacket' that has a Hillcrest or equivalent sensor at the mid-point of all upper body joints (one or two on your spine, in-between your hips and neck, one as close to the center of your eyes as possible, one at each shoulder rotator cuff, mid upper arm, mid forearm, etc) and to use the combined data from those sensors to figure out exactly where all your body parts are with respect to each other using basic trig. I would imagine that some level of calibration would need to be done that involves taking physical measurements of all your limbs (for the purpose of determining where the sensors are WRT each other), and then mapping those values into the game's measurement system.

This setup would be for games where you're seated, but the concept could easily be extended to standing games by adding more sensors below the waist. Doing so would provide full body position determination WRT a fixed point (namely, your feet). The only movement that would require other means to detect would be jumping (no sensor angle would change during a jump)

This seems like such an obvious solution that it has to have been considered and discarded for some reason.

Thoughts?

You can only really measure the movements of rigid bones with that sort of orientation-only setup. With the spine, and especially the neck, you have to try and fix a lot of sensors in known fixed positions relative to vertibrae in order to account for the ability to shift the head sideways (i.e. translation without rotation).

Like this? http://www.youtube.com/watch?v=MSTge5IDxF4

[Edit] I have been beaten to posting my own stuff X.x

The same sensors can detect jumping, as long as you assume a ground plane is where the feet are at rest. Positional tracking is good for a few seconds at a time.

Also if building the wired version the cost goes down by a good amount.

Fredz wrote:

Mel wrote:This seems like such an obvious solution that it has to have been considered and discarded for some reason. Thoughts?

Answer : cost ($4000 for full body tracking).
See : http://www.mtbs3d.com/phpBB/viewtopic.php?f=120&t=15564
Would be less expensive for just the upper body, but still nowhere near a consumer budget.

That's precisely what I had in mind. Glad to see someone has done it. But yes, 4 large is a pretty big chunk of change. Maybe wide-spread Rift adoption will bring that price down.

RoadKillGrill wrote:Like this? http://www.youtube.com/watch?v=MSTge5IDxF4

[Edit] I have been beaten to posting my own stuff X.x

The same sensors can detect jumping, as long as you assume a ground plane is where the feet are at rest. Positional tracking is good for a few seconds at a time.

Also if building the wired version the cost goes down by a good amount.

You're the guy in the video? Great stuff. I was envisioning a wired version myself, or at least all sensors wired together with a central wireless 'master'. This could easily be stitched into a jacket of some kind.

RoadKillGrill wrote:Like this? http://www.youtube.com/watch?v=MSTge5IDxF4

[Edit] I have been beaten to posting my own stuff X.x

The same sensors can detect jumping, as long as you assume a ground plane is where the feet are at rest. Positional tracking is good for a few seconds at a time.

Also if building the wired version the cost goes down by a good amount.

Do/can you sell an upper-body only, fully wired version of this system?

@bobjwatts: The video is on YouTube3D. It is not just anaglyph (I think thats the default, though). If you click the 3D button on the lower-right you will get more options for side-by-side, interleaved, checkerboard, etc. The Rift itself is side-by-side parallel like in the first video I posted.

@donkaradiablo: My driver does not support DOOM3. Anyway, my computer is fast enough to play and record in 3D, its just that the tracking software has a problem with this.

@Quant: The Perception driver already supports checkerboard. Nvidia 3D Vision output could be an option down the road. If you are suggesting I somehow hack the Nvidia drivers to support the Rift, I don't think that's going to happen.

@BlackShark: I understand where you are coming from with the frame delay thing, and I totally agree. However there is a limited amount of time before the Rift dev-kits ship and I am trying my best to have something (anything) ready for December. So in that interest I have had to make compromises and, in some cases, hack things together just to get them to work. I hope to improve things over time.

The distortion when using the camera roll is something I am aware of, I'm just not sure if its an easy fix. I will continue to look into it and see if there is a solution. I have also noticed this on the Vuzix headsets with certain games (like UT2004) so maybe it is not so simple. I imagine if you are working on a custom game (or have the source code) this wouldn't be an issue, but you only have so much control by intercepting DirectX.

@brantlew: Neck modelling sounds cool, and its certainly something I'd like to explore. But again, there is not a lot of time before December so I just need to focus on getting things working proper without too much fancy business.

Hey Cyber,
When did you get in for a RIFT Kit..Pre Kickstarter, $275.00 November Kickstarter or December Kickstarter ?
Only asking, because I was wondering, when you will have one in your hands (in/on your face) to work on your driver ?

Not sure which wave he was in, if i remember correctly he has 2 going his way. Not sure if he got in on the pre-kickstarter + kick starter or what combination hes got

Will you add a feature to your driver that allows people to feed in their own tracking data?

And also a way for games to opt out of using your driver if they are already rendering for the Rift natively (so we don't get a quadrupled doubly warped image)? Preferably a way that will work on all Rift drivers (you should discuss it with the other vendors).

@3dvison: I will be getting my Rift in December I guess (unless someone wants to send me one early...). I am trying to see if I can at least get the tracker soon, we will see.

@2EyeGuy: Yes, I plan to have a plug-in type interface where people could add support for their own tracking devices. I don't have a lot of time before December, so I can't promise it will be in there at launch. But its in the pipeline.

If you want to turn the software off you can simply disable it via the GUI, or just close the window and it will stop working. Pretty simple. Also, it only works for the small number of games with profiles, it won't just hook into any game.

I would swap with ya Cyber, but I'm December also.

cybereality wrote:@donkaradiablo: My driver does not support DOOM3.

Maybe this would help:
https://developer.valvesoftware.com/wik ... ding_Tools

How about the idea of TF2 support and using Source FilmMaker? (leading to much better looking Vireio Perception videos maybe )

Nice work cyber!
I will certainly buy your driver as soon as it will be available.

I would love to see Racedriver Grid supported, and Hunted: The Demon Forge could be a good 3rd person game in VR.

OmniAtlas wrote:Flight Simulator X support Please!

I secound that

It would be nice to have some kind of support for Lord of the Rings Online. Not to "play" the game, do quests and level, but just to wander around the wonderful landscapes. The game has an option to hide all the interface and, with the right mouse button, you can look around, so it would be awesome just to walk around Middle Earth. Oh, also pressing * switches to first person mode. Wandering around a virtual Middle Earth would be a dream come true. By the way, the game supports from DX9 to DX11, so that would be no problem for Vireio.

The landscapes of that game are so huge and so beautiful, it would be a perfect showcase for the rift.

Hey Cyber. Steam has some pretty awesome deals happening right now in their Halloween sale. Will your driver work with Steam versions of games? I don't think you officially support any of the sale items, but I might pick some up just in case. Bioshock 2 for 5 bucks? Can't lose.

Isn't Cyber a City mouse ? A NY City one at that..Stay safe Cyber.

@cyber - So are you on the east cost? Everything alright?

Mel wrote:Hey Cyber. Steam has some pretty awesome deals happening right now in their Halloween sale. Will your driver work with Steam versions of games? I don't think you officially support any of the sale items, but I might pick some up just in case. Bioshock 2 for 5 bucks? Can't lose.

Ohh, bioshock 2 support if only for the bit where you walk outside for the first time. So cool.

Dead Space is on sale on Steam right now.
The lack of a HUD would make the game optimal for VR.

Valez wrote:Dead Space is on sale on Steam right now.
The lack of a HUD would make the game optimal for VR.

But, as you are aware, the Dead Space series is in third person. 3rd person cameras swing around a very large radius (sorry if this is the incorrect term) that wouldn't at all be suited for an HMD-only setup. The radius is smaller when aiming, but still some distance away from a first person view, and thus similar problems. Maybe a first person camera hack would be possible, but I see no modding community around the Dead Space games (thus it is likely extremely difficult to mod), so I almost completely doubt that that will ever happen. Might as well focus on first person games, although I would love to see support for GTAIV - there is a first person camera mod, and I love to walk around the city on foot in first person.

You probably meant playing the game without head tracking, though. That could be fun. I wonder about motion sickness from swinging the camera around, though. I'm personally looking forward to testing that kind of thing out myself (at least for a couple of seconds, one single time ).

I picked up a few games on the sale. Namely:

Amnesia, I Am Alive, and Condemned

And don't worry, I'm totally safe. The area where I live wasn't really hit bad at all. Just stuck in the house (not a bad thing).

Funny how those game titles describe your situation...

Ha!!!

cybereality wrote:I picked up a few games on the sale. Namely:

Amnesia, I Am Alive, and Condemned

And don't worry, I'm totally safe. The area where I live wasn't really hit bad at all. Just stuck in the house (not a bad thing).

Amnesia would make an incredible VR experience. I think the game would be very well suited for Rift. I believe it also has quite a lot of community-created content so it offers longevity too. Something you might want to consider.

jf031 wrote:

Valez wrote:Dead Space is on sale on Steam right now.
The lack of a HUD would make the game optimal for VR.

But, as you are aware, the Dead Space series is in third person. 3rd person cameras swing around a very large radius (sorry if this is the incorrect term) that wouldn't at all be suited for an HMD-only setup. The radius is smaller when aiming, but still some distance away from a first person view, and thus similar problems. Maybe a first person camera hack would be possible, but I see no modding community around the Dead Space games (thus it is likely extremely difficult to mod), so I almost completely doubt that that will ever happen. Might as well focus on first person games, although I would love to see support for GTAIV - there is a first person camera mod, and I love to walk around the city on foot in first person.

You probably meant playing the game without head tracking, though. That could be fun. I wonder about motion sickness from swinging the camera around, though. I'm personally looking forward to testing that kind of thing out myself (at least for a couple of seconds, one single time ).

Yes, this could be a problem. But I wonder if this would be really a big problem. And I think Palmer or Dycus mentioned 3rd person games as a viable option for VR. And it would be at least worth the try and the experience.

jf031 wrote:But, as you are aware, the Dead Space series is in third person.

Dead Space Extraction is in First Person.

I've played Tomb Raider Anniversary in third Person VR with head tracking making your orbit around the character. I thought it was fine. There are also other ways you could do third person in VR. But first person hacks for third person games would be good too.

"A man wearing nothing but his boxer shorts and a RIFT head mounted display, unknowingly wandered out into a hurricane today"

Good to hear you're OK Cyber..I knew that could not be you, you're an old VR veteran, you would never make that rookie mistake..Well not twice anyway...HaHa

This Amnesia game sounds cool, will your driver be supporting it ?

3dvison wrote:This Amnesia game sounds cool, will your driver be supporting it ?

I would like too. Still have yet to play it yet.