Meant to be Seen

yes, it's the projection for wide FOV that results in too much stretching on the edges of the screens

foisi wrote:yes, it's the projection for wide FOV that results in too much stretching on the edges of the screens

So, you would basically reduce your frame rate by 60-70% in order to render to the HMD?

mmh I don't think that the number of camera affects the framerate that much

foisi wrote:mmh I don't think that the number of camera affects the framerate that much

But, don't you need to render the scene completely from a different angle? You're using the 3 camera views at the same time, aren't you?

i'd suspect each camera to cover 1/3 of the FOV, and everything is rendered in single rendering pass.

But then - would a single-display setup too require a multi-camera driver? Wouldn't iz3d work with it properly?

This is what is done in my demo :
graziano you're right, 1/3 of the view is covered by one camera (for each screen)

(10000 views \o/)

foisi wrote:(10000 views \o/)

Congrats!

Any news?

Thanks Zoide

Yes I'm still working on version 2 when I find some time, but nothing I can show here for now (maybe in a few weeks).

Hello Foisi,

I have ordered a Rift last month. Funny enough I though of the dual display for extra resolution not so much for the 180 FOV. Then walla!!!...here you are with the IfE.

I am on the designing stages of VR Game although I not a gamer, i come from the design/Animation professional sector. Currently checking different engines, UDK, Cryengine, Unigine. Honestly speaking a lot of the tech talk here is out somewhat over my head. I am technical incline but new to this type of technology.

I will be looking for the components to build a unit or two based on your designs.
This would be primary to help me develop my VR game.

My concern is been able to use it with any existing SDK? Can you chime in on this matter since i keep on reading that there is no software to use it with. maybe something can be worked out?

Please advice

Warmest Regards!

Hi Hannibalj2,
you're right there are no software to use with the InfinitEye for now as I am focusing on hardware until I finish the second prototype. What matters for resolution is the angular resolution which is approximately 7.1 pixels/degree on the Rift (640px/90°) and with the InfinitEye it should be the same (maybe a little more because the lenses have a longer focal length but certainly not double).

Thanks for the response. So anyone that builds a DIY HMD is out of luck when wanting to use it for development?
Is the incompatibility due to the hardware components? Could there be a possibility of using the UDK sdk if the right hardware components are used, ei. Tracker, displays, etc..?

Thanks for your time

foisi wrote:@TheLostBrain: Here are some exported images from my app, I will try your solution this weekend if I find the time

p1.png

p2.png

p3.png

p4.png

p5.png

@Fredz: cool, I will make something viewable on the infinitEye from the 360° stereo images, thanks for the link

@ripcurl123:
I don't know if they are the best fresnel lenses but they are the best I could find (a 60mm focal length fresnel lens with approximately the same size could ideally replace the 2 stacked 120mm focal lenses (and with a smaller pitch it will be even better))
I have ordered something like a dozen different lenses (including these 120mm focal length fresnel lenses and 180 mm focal ones) because after I completed my previous project (a "rift-like" ancestor hmd) I wanted to create something with a FOV as big as I could and fresnels were the only lenses that provided it without being too heavy and also being easily cuttable compared to glass lenses.
to answer your other question : yes the grooves are facing the screen and the lenses are stacked with no spacing between them

Edit: thanks to mediavr for the pics, here is a normally viewable (not tested yet) version of your belmorepark pictures

belmorepark1InfinitEye.png

I tried this configuration ten years ago! You have to set up with FISHEYE. Else the image will be spread larger on top and at the bottom! The configuration of these pictures are "Mercator" panorama's, which are exellent with curved screens or projection screens. Try also "STEREOGRAPHIC" but Fisheye is ok. It would be the same configuration as the Oculus rift but Full fisheye = 180 deg (Oculus is 100 deg or ~ 60 procent of the fisheye image) Notice that the middle of the two fisheyes would be cropped in the middle for the nose. thumbs up

Hey Foisi, really impressive prototype you have there!

I'd love to shoot some video footage for you with my fisheye action cam rig (head mounted or handheld), but their FOV is only 140 degrees.

So you need three cameras to cover 180? In theory I could get a third Braun Sixzero, or maybe slap one between my two VIO POV.HDs, on top of that "silicone ears" microphone, but ... I have no idea how the footage should be combined into a single file that would be optimally viewable with the InfinitEye. Also, the middle camera would be different from the side cameras, so I'm not sure how much incompatibility there may be between the video clips.

Or is it enough to just use two cameras that diverge outwards (with a bit of overlap in the middle), to cover the needed 180 degrees?

Well, I did the simplest, quickest thing, and pointed my Sixzeros outwards on top of the Free Space:



They haven't been placed with 100% engineering precision, but I'm trusting that small alignment errors can be fixed in post edit.

Would you like to me to try shoot and edit some test footage for you with this rig?

Here's a bunch of test snapshots I took with the above rig, and placed side by side:

http://imgur.com/a/uk0kI

I didn't do any specific alignment, because I don't exactly know how. But you know, just to get an idea... I know their aspect ratio does not match InfinitEye, but this is easily fixed by cropping.

Skaven252 wrote:Here's a bunch of test snapshots I took with the above rig, and placed side by side:

http://imgur.com/a/uk0kI

I didn't do any specific alignment, because I don't exactly know how. But you know, just to get an idea... I know their aspect ratio does not match InfinitEye, but this is easily fixed by cropping.

You can look at the alignment of the freshnel lenses on the construction pics. Hopely it's work. Also the lens distortion is very important to get the perfect 3D pairs.
I don't know the angle of the camera's but it seems to be large enough to get 180 degrees together.

I was looking through one of my Maker Magazine articles the other day & ran into a page that explains how to make a lens from gelatine... cheap & very effective for lens prototyping.

3dpmaster wrote: I don't know the angle of the camera's but it seems to be large enough to get 180 degrees together.

Looks like not quite... a horizontal object like the balcony railing can be used to tell it's not quite 180.

I wonder how much can be compensated by editing, cropping and offsetting though, maybe there's only a bit of loss in the outer edges. Better than nothing. I haven't tried this with my VIO cameras yet. They have a bit wider FOV I think, and thanks to the way the mounts are shaped it's easier to point them outwards (the Sixzeros are jammed against each other and can't be pointed further apart than that).

Maybe some other camera make and model has an even wider FOV.

Hi, I will try your stereo pairs when I will finally be able to finish the second prototype (I'm waiting for parts to be shipped) hopefully before the end of june
Even if the total HFOV is not 180° on them, I'm sure I can do something to be able to view them properly
Thanks

why does the InfiniteEye have only 63 deg overlap. if I see the optics, the overlap should be bigger?

Yes you're right. The horizontal FoV is actually > 180° (something like 210°) and the stereoscopic overlap is about 90°

Hey

I made a new thread for the V2 prototype so I will be posting progress there from now http://www.mtbs3d.com/phpBB/viewtopic.php?f=138&t=18369

Instead of using a mini display port to dual hdmi adapter, couldn't you just use 2 hdmi cables and set up a dual monitor eyefinity? Seems to me like that might be the easier route.

Can i use 2 galaxy s2's as display for starters?

I have no experience in dealing with lenses, so i dont want to invest so much money into displays in case i fail to build one.

I can develop a demo game in Unity to render 3d view onto 2 smartphones.

hi, i am very impressed by the specs! Is it correct that the real stereoscopic fov is smaller than the one of the rift?

Hello!
I want to ask you how you added head tracking using YEI 3 space sensor embedded in Tuscany?

foisi wrote:
- Horizontal FOV = 180° (calculated)
- Horizontal stereoscopic FOV = 63° (calculated)
- Vertical FOV = (to be determined)

Hello!

I've got a question about the FOV; for a school project I'm working on an HMD and I am wondering how do you calculate the horizontal/vertical FOV? Perhaps someone can help me in this?

Thanks!

If you are trying to figure how to calculate the Horizontal and Vertical when given the diagional, you can do that using trigonometry. But, perhaps you are wanting to figure the possible field of vision knowing only the geometry of the HMD and the lens info.

I have made a rough estimate by assuming that there is no lens, and that the iris of the eye is the apex of a triangle.

[img][/img]

Note that you have to divide the display into two parts by drawing line perpendicular to the screen, to the iris. Then, calculate the two areas separately.

But, I think the strong lenses that are typically used will have a significant influence on the angles. I don't know how to manage those lens calculations. In that case, if the lenses are available, actual measurements can be made. Put some graph paper where the display will be. Then, with the paper mounted at the correct distance, and with a pencil, mark the boundaries of your vision on the paper itself. Then you can draw the display screen on the graph paper and use the Pythagorean theory to calculate angles.

-Joe Dunfee

schipleon wrote:I've got a question about the FOV; for a school project I'm working on an HMD and I am wondering how do you calculate the horizontal/vertical FOV? Perhaps someone can help me in this?

Cadcoke5 gave the basic principle but as he said you'd need to know the lenses characteristics if you want to calculate the FOV accurately. That's quite complicated for aspheric lenses and even more for stacked fresnel lenses because you'd need to know the refractive index of the material used and the equation of the surfaces of the lenses. These are generally not available for off-the-shelf lenses and should be quite hard to evaluate experimentaly.

Another option would be to use a simple laser pointer and some sort of millimeter paper to physically calculate the FOV as show in the photo below. But you'll not have a unique FOV since it depends on the distance between the eye and the lens. Also before proceeding you'd need to know at which distance the lens must be placed from the display in order to have the eyes focus at infinity.
And to have a meaningful FOV calculation you should also use a sensible value for the center of projection in the eye, which is not on the surface of the cornea and neither at the pupil position or the center of rotation of the eye according to what I've read (and hopefully understood). In this paper (schema below) the center of projection is shown to be at the position of the nodal point in the Listing schematic eye model.

Thank you all very very much for the help! this is very helpfull!

Regards!

Thanks for the link about the eye. It was a very nicely done summary of a lot of technical info.
http://www.oculist.net/downaton502/prof ... .html#redu

I did note one typo about stereoscopic vision. The site said, "...beyond about 2000 feet stereoscopic judgments are of little value." I think the 2000 feet should have been 20 feet. Other sources generally say that at beyond 18 feet, stereo separation is of little value. I think many years ago, I read one source saying 15 feet as the practical limit.

-Joe Dunfee

Not sure if it's a typo, found this quote in a paper I can't seem to find anymore :

In central vision, under favorable conditions and for observers with excellent stereoacuity (12 sec of visual arc), stereopsis (a binocular depth cue), can provide a meaningful cue to distances up to ~450 m. In the presence of both binocular and monocular cues, stereopsis is the most effective distance cue in central vision up to ~65m. In peripheral vision, stereopsis is less effective, and motion parallax, a monocular cue, may be more effective.

The ~450m distance gives a ~1500 feet distance, not that far from that 2000 feet quote. But that's when there are no other depth cues available. I've seen that ~65m figure (~213 feet) mentioned in several other papers in real life situations when other depth cues are more prevalent.

An illustration of this (not a linear scale) :

Can you post modified VireIO driver, please?

foisi,
as you know I'm working on my own hmd design, and I´ve been experimenting with the fresnels from optolife. I´ve noticed the image quality quickly deteriorates as i move my pupils away from the optical center of the fresnels. Is this a problem with infiniteye? What IPD range will be supported?

adventuresinz wrote:foisi,
as you know I'm working on my own hmd design, and I´ve been experimenting with the fresnels from optolife. I´ve noticed the image quality quickly deteriorates as i move my pupils away from the optical center of the fresnels. Is this a problem with infiniteye? What IPD range will be supported?

Did you try flipping the lenses over? The off-axis optical quality (and distortion) depend on which side of the lens you view through. Of course, the current SDK software does not correct for off-axis (tangential) distortion, so you would need software for that.

geekmaster wrote:Did you try flipping the lenses over? The off-axis optical quality (and distortion) depend on which side of the lens you view through. Of course, the current SDK software does not correct for off-axis (tangential) distortion, so you would need software for that.

I've tried all 4 combinations with 2 stacked fresnels (120mm focal length each). The stacking might be a problem. Having both fresnels with even/flat side towards eye seems to work best, at least for me, but image-quality and brightness gets worse as i move my eye away from the center when fresnels and display are fixed. At 1 cm off axis it´s bad - not so much at 0.5. So +/- 1 cm IPD (+/- 0.5 cm off axis for each eye) might be tolerable but far from ideal and significantly worse compared to regular lenses. On axis it works great, not much different from regular lenses.

How do you handle chromatic aberration? In my tests I´ve noticed significant chromatic abberation at the pheripheral and at the area right in front of my nose.

adventuresinz wrote:

geekmaster wrote:Did you try flipping the lenses over? The off-axis optical quality (and distortion) depend on which side of the lens you view through. Of course, the current SDK software does not correct for off-axis (tangential) distortion, so you would need software for that.

I've tried all 4 combinations with 2 stacked fresnels (120mm focal length each). The stacking might be a problem. Having both fresnels with even/flat side towards eye seems to work best, at least for me, but image-quality and brightness gets worse as i move my eye away from the center when fresnels and display are fixed. At 1 cm off axis it´s bad - not so much at 0.5. So +/- 1 cm IPD (+/- 0.5 cm off axis for each eye) might be tolerable but far from ideal and significantly worse compared to regular lenses. On axis it works great, not much different from regular lenses.

How do you handle chromatic aberration? In my tests I´ve noticed significant chromatic abberation at the pheripheral and at the area right in front of my nose.

Off-axis CA is significant, and would require a software adjustment. However, that can only be corrected to the nearest sub-pixel (horizontally). For good stereoscopic alignment (no vertical parallax), using offset lenses requires tangential distortion (as formulated at the wikipedia page on lens distortion correction, linked in many of my other posts in these forums).

Hi,

I wonder if we assembled the HMD according the first page specifications how would we try it on HL2 for instance?
Do we need just some game adjustments to start using it or we need a special driver? Can somebody clarify that?
Thanks in advance and sorry for poor English.

dkd wrote:Hi,

I wonder if we assembled the HMD according the first page specifications how would we try it on HL2 for instance?
Do we need just some game adjustmens to start using it or we need a special driver? Can somebody clarify that?
Thanks in advance and sorry for poor English.

You need special custom image warping (different from what the Rift uses), and you need to display the stereoscopic pair on images on separate monitors. You would probably need some sort of driver or other game customization to support this (perhaps a modified version of the Vireio Perception drivers).