It's look very good on Oculus Rift, but onestly, looks like a normal spherical
camera without 3D effect.

Look at this ODS example (rendered with Blender):
https://www.clodo.it/files/projects/povray_ods/gooseberry_benchmark_panorama.jpg

Look how the left and right images are different:
https://www.clodo.it/files/projects/povray_ods/blender_diff_ex.png

Now, look your Cave left and right compare:
https://www.clodo.it/files/projects/povray_ods/cave_diff_ex.png

I think rocks are too far.

Dark scene are very good for HMD.

Also about the resolution:
8192 x 2048 side-by-side can be acceptable with toys like Cardboard & Android.
But with HMD like Oculus Rift or HTC Vive, it's too low.

Look this comment:
https://forums.oculus.com/viewtopic.php?f=20&t=30852



> You need this value of pixel coverage or "pixel density" or
> "pixel per display pixel" or "eye buffer scaling" as it's variously called.

> to cover one monoscopic turn.

It mean 6000x6000 in top/bottom, or 12000/3000 in side-by-side.
At least for GearVR, i don't actually know the correct value for other HMD.

Apart the rendering time, rendering images for HMD it's not an issue.
The issue i'm studying it's about video/animation:
this kind of resolution are outside any H264 or HEVC/H265 resolutions,
and other types of codec are not hardware accelerated,
and Oculus Rift and HTC Vive want 90 FPS, Playstation VR want 120 FPS....

----------------
I try to render the plants_demo of Gilles Tran, but i'm not satisfied.
There are a lot of flying object like butterfly and leap, and
i currently don't find a good position for a good 3D effect.
And it's too light, HMD use OLED display at few millimeters to eyes, it's better
dark scene like Cave.

Some ODS render 6480 x 6480 pixels, Top-Bottom
(published here: https://forums.oculus.com/viewtopic.php?f=28&t=30854)

Mirrors - I made this
https://www.clodo.it/files/projects/povray_ods/mirrors_20160316.jpg

Stacker Day - POV-Ray sample scene adapted for ODS (i change some reflection and
camera position)
https://www.clodo.it/files/projects/povray_ods/stackerday_20160316.jpg

Fractals 1 - POV-Ray sample scene adapted for ODS (i change ball position and
camera position)
https://www.clodo.it/files/projects/povray_ods/fractals1_20160316.jpg

Fractals 2 - POV-Ray sample scene adapted for ODS (i change only camera
position)
https://www.clodo.it/files/projects/povray_ods/fractals2_20160316.jpg

Wineglass - POV-Ray sample scene adapted for ODS (nothing changed. Apart camera
type of course)
https://www.clodo.it/files/projects/povray_ods/wineglass_20160316.jpg


----------------
I released on GitHub a fork of POV-Ray.
https://github.com/Clodo76/povray

Basically, ODS with IPD=0 are the same render as spherical camera.
So, i added two option to the current spherical camera: 'ods' (default 0, for
backward compatibility) and 'ipd' (default 0.065)

ods = 0, original POV-Ray spherical algorithm.
ods = 1, ODS, only left eye
ods = 2, ODS, only right eye
ods = 3, ODS, side by side
ods = 4, ODS, top bottom

Using user_defined camera or this patched version is the same.
As Chris said, C version is little more faster, but only in very very very
simple scene. Irrilevant imho.
I published only if someone want to continue experiments, because it's more
readable in C rather than the 6 user_defined functions.
And also because i write that before the Chris user_defined implementation :P
----------------
Soon i will create a POV-Ray Wiki page about ODS.

Ciao!

Post a reply to this message