[Nova66]

Contact them directly via the website

I might give that a go Seghier, thanks. To be honest I didn't even think of that because after all we're here in a Beta forum talking about Beta issues, surely everyone would benefit from that discussion. I know flame wares get tiresome but reducing communication to a trickle can't be the answer either


Remaining hopeful,
Andrew.

[seghier]

i hope they fix that soon ; maybe because it is new feature and need more tests

[juan]

When I look at stereo panoramas created with V-Ray or OctaneRender, they handle the need for NO parallax at the poles perfectly well but Maxwell introduces counter rotating warped spirals at the zenith & nadir which makes it impossible for your eyes to merge the two Left/Right eye images as your gaze moves towards either of the two poles.

Hi Andrew,

We are in contact with the creators of the lat-long lens model and they believe (as we do) you are getting that effect because you are not using a stereoscopic separation map to reduce the stereo effect in the zenith and nadir zones. We are reviewing this whole area to see how it can be improved and if there is something that can be improved here we will do our best to fix that. Thanks a lot for your feedback.

[Nova66]

Thanks Juan for taking a look at this issue, I'm so excited to get this thing working properly for me.

We are in contact with the creators of the lat-long lens model and they believe (as we do) you are getting that effect because you are not using a stereoscopic separation map to reduce the stereo effect in the zenith and nadir zones.

I totally agree with this analysis and if you take another look at my posts in this thread you'll see that I've been trying to use the "Separation Map" option but I can't seem to make it work. The help system doesn't give any clues as to how it should be set up so at this stage I don't know if I'm doing something wrong or it simply does not work. I've tried all sorts of grey scale gradients and even looked online at how other applications define the Separation Map and nothing seems to make a difference. I really think Maxwell Studio's implementation of the Separation Map is broken.

Can you please give us some guidance as to how the Separation Map is to be used and if you can make it work at your end?


Thanks,
Andrew.

[seghier]

Ive just tried Maxwell v3.1.99.10 on the sample file that I provided above and even though there now is some parallax at the equator, the two poles are even more warped than they were before, when they really should have NO parallax at all. I'm using the "PSViewer 2" app on my Android phone and the poles have an extreme spiral in them and are so far apart for the left and right eyes that it's impossible to make the two images converge.

When I look at stereo panoramas done with Octane Render and now Vray, the stereo effect works really well and the poles don't cause any convergence problems at all. There is something still very wrong with Maxwell's implementation of the Stereo Panorama lens

I tried different values for the "Parallax Distance" but nothing seemed to converging on a usable solution.

From what I've read about Andrew Hazelden's work elsewhere, it feel like the "Separation Map" is the magic ingredient to progressively reduce the amount of parallax as you approach the poles but no matter what kind of grey scale gradient I try to plug into that slot, it seems to make no difference to the output at all.

What kind of image are we supposed to use for the "Separation Map" and how are you supposed to make it work. The online help doesn't seem to offer any clues at all.

Am I wrong in thinking that a correct solution should look something like this mockup which doesn't have any warping at the poles?


Cheers,
Andrew.

Hello Andrew
i just want ask you ; why you use separation = 6.4 cm ? i don't understand the problem ; can you show a right result created with other render engine ?

[seghier]

i made two tests

[Nova66]

i just want ask you ; why you use separation = 6.4 cm ? i don't understand the problem

Hi Seghier, the pupils in my eyes are 6.4cm apart so that's where that number comes from, I think the default value is 6.0cm. If you set the Separation value to 0.0cm then you don't get any parallax and the stereo effect won't work. On the other hand, if you set the Separation value too high, like 50cm, the world would seem very small and it would be as though you were a giant looking at it.

can you show a right result created with other render engine ?

I don't use Vray or Octane so I can't show you a direct comparison, the best I can do is to show that mock-up that I included in my previous posts. Additionally, when I download stereo panoramas that other people have done using either Vray or Octane, the zenith and nadir are perfectly viewable whereas the images that you get from Maxwell do that exaggerated warping at the poles that you see in your tests.

As Juan said, the problem with these stereo images lies with the fact that we're not using a Separation Map. But my problem is that I can't make the Separation Map option work.

At the end of the day, I'm using Google Cardboard and an app on my Android phone called "PSViewer 2" to view the stereo panorama images. I'm hoping this will be a compelling way to show clients the 3D space that is being designed for them.


Cheers,
Andrew.

[seghier]

Thanks Andrew
i used different separation maps but i didn't see any changes

[seghier]

i made one test with 2 cubes and one with a sphere ; i think it's logic to get the first result with sphere ( like in your scene )

[juan]

Hi guys,

Below you can see docs & tips about these lenses. We are talking to this guys to update these docs with Maxwell specific info (and our online documentation too, of course)


Fisheye Stereo Controls
https://github.com/zicher3d-org/domemas ... r-controls

LatLong Stereo Controls
https://github.com/zicher3d-org/domemas ... r-controls


Tips & Tricks Wiki Topics of Note

Description of Stereo Control Textures
https://github.com/zicher3d-org/domemas ... o-textures

Controlling Stereoscopic Depth
https://github.com/zicher3d-org/domemas ... opic-depth

Stereoscopic Viewer Comfort
https://github.com/zicher3d-org/domemas ... er-comfort

Adjusting the Stereobase
https://github.com/zicher3d-org/domemas ... stereobase

Hyperstereo
https://github.com/zicher3d-org/domemas ... yperstereo

[seghier]

Thanks Juan
also in the website of the developer
http://www.andrewhazelden.com/blog/2014 ... 1-6-alpha/
every software have different settings

[Nova66]

Below you can see docs & tips about these lenses.
We are talking to this guys to update these docs with Maxwell specific info

Thanks for the link Juan and it's great to hear that Maxwell's documentation will be updated to reflect this information.

And now for the bad news, the Separation Map that's provided with that documentation does not work with Maxwell. From the screen capture of the Maya "LatLong Stereo Shader Controls", I can see that there is something called "Separation Multiplier" which is a value between 0.0 & 1.0 and has a texture channel associated with it.

Maxwell's interface only has the Separation Map option but no Multiplier value. Do you think that this multiplier value has not been plugged into Maxwell's code and that is why the Separation Map texture has no effect on Maxwell's output?

The Separation Map that's provided in the documentation looks like this:

This was totally non intuitive to me and I've got to say that I never would have guess it's format.

On a related note, Andrew Hazelden's Blog says that the Separation Map should look like this:

This is more along the lines of what I was expecting but now I'm unsure as to which help file to believe.


Cheers,
Andrew.

[Dover Studios]

Hi this is Andrew Hazelden.

I'm one of the co-developers of the Domemaster3D shader. I'd like to apologise for not being more active in this forum. Juan Canada and Maxwell have been very nice to the Domemaster3D project over the last year+ and really supportive of the project's efforts.

I haven't been able to be involved with the Maxwell lens shader port as much as I would have liked to over the last few months since I've been working fulltime on my own projects that involve the development of VR previz, and panoramic 360 degree compositing tools through my blog. Since I didn't test the most recent 3.2 builds of the Maxwell ported stereo lens shaders I am partially responsible for the rendering issues you have discussed on this forum not being resolved today. For that I am sorry.

* * *

Today I created a Maxwell Studio .mxs version of the standard Maya/Domemaster3D mental ray scene called "LatLong_Stereo_CubeX".

Mental Ray CubeX LatLong Stereo Demo Scene
Here is a stereoscopic 3D LatLong animation of what the mental ray CubeX scene looks like:
https://www.vrideo.com/watch/bavxEI6m


Maxwell CubeX Scene Download
You can download the Maxwell CubeX scene files here to do your own stereo tests. You can also check out the sample CubeX 3840x1920px Latlong Stereo images that are included with the scene:
http://www.andrewhazelden.com/projects/ ... _cubex.zip

Tip: If you want to have the scene render faster, you can try lowering the SL quality setting and the resolution of the cubeX demo scene before rendering the .mxs scene file.

Maxwell Studio CubeX Screenshot


Maxwell Rendered LatLong Stereo Image
(The left and right views were merged to red/cyan anaglyph for inline forum based web viewing)


* * *

Notes on the Current LatLong Stereo Shader in Maxwell 3.2 as of Today

In Maxwell Studio if you change the separation value to 6.5, and the Parallax Distance to 3600 you should get a better result with the LatLong Stereo shader and a closer match to the original Domemaster3D shader's Maya/mental ray based default settings of camera separation 6.5 cm / Parallax Distance 360 cm.


The issues you have been mentioning in this forum with the first Maxwell integration of the stereo lens shaders should be able to be fully resolved shortly and come down to two things:

Issue Part 1 - Lens Shader Measurement Scale

Maxwell Studio and .mxs files use meters internally for the scene scale. The Maxwell integration plugins that work with Maya/Max/etc... all use a scene scale translation phase.

It was initially discussed that it would seem logical to make the camera separation value work in centimeters for consistency across each of the Maxwell integration plugins and Maxwell Studio since everyone working in stereo 3D is used to entering a roughly 6 to 6.5 cm stereo camera separation value / IPD. This is where part of the confusion starts since not all parts of the lens shader math did the scene scale conversion the same way and the current GUI didn't implicitly state the units of measure on every one of the lens shader GUI's input fields (mainly parallax distance for example).

Today I've been in discussion with Juan over theses issues and I put together an updated Maxwell XUI file for the lens shaders that should make things clearer for the next Maxwell Render release that will hopefully be available shortly.

Issue Part 2 - Control Maps

There is a separation map feature in the Maxwell Studio "Lat-Long Stereo" shader, and a separation/turn/head tilt map feature in the "Fisheye Stereo" shader that are designed to use the same stereo control texture maps as the original Domemaster3D shaders do. If the scene scale is not adjusted correctly in "part 1" above then the control texture map can't really take effect properly in "part 2".

The stereo control texture maps are typically an RGB image (like a PNG texture) that have a black to white color in the image which represents the area of the rendered frame that will have full stereo effect or a reduced stereo effect. The control texture maps take care of "swirly" artifacts that would occur in the pole zones during the turning "panoramic" stereo camera rendering phase that is used to make a continuous and seamless 360 degree stereo effect.

The control maps are applied by the Maxwell lens shaders using "screen space" coordinates which means they are directly mapped 1:1 with the areas in the final rendered image. In a separation texture map pure black means no stereo effect will appear in that area, and white means the stereo effect will be at full strength. The areas of shading between pure black and white have a feathered out stereo effect to gradually reduce the stereo depth. For a latlong stereo control map, it really is just a simple gradient that chooses where in the image you want to start fading out the stereo effect.

The sample control texture maps files that are included with the regular Domemaster3D shaders are installed here:
C:\Program Files\Domemaster3D\sourceimages\

The texture "latlong_separation_map.png" is for feathering out the stereo effect in the pole regions in a latlong rendering. The zenith or upwards sky facing pole region is at the top of the texture map, and the nadir or downwards facing pole region is on the bottom of the texture map.

latlong_separation_map.png


The textures "separation_map.png", "turn_map.png", and "head_tilt_map.png" are for use with 180 degree angular fisheye renderings which are also called domemaster frames due to their use in planetarium "dome" theaters. These 180 degree angular fisheye images typically have a black circular mask applied to the perimeter of the frame so anything outside the circular area in the final rendering is cut off.

The 180 degree angular fisheye separation map texture may look a bit strange but what it is indicating is the area just behind the viewer's head when seated in a dome theater has the stereo effect reduced where the viewer may be fully tilted back in their chair and looking back past the top pole region of the theater. The head tilt map has two shaded areas that are used to cause the stereo rendering settings in the lens shader to adjust for the fact a seated viewer in the planetarium theater is rolling their head at a slightly tilted angle when looking to the far left or far right area on the "front half" of the dome screen.

separation_map.png


turn_map.png


head_tilt_map.png


As a note, a screen space control texture map is typically painted in Photoshop using a layer placed ontop of one of your own panoramic renderings. This helps make the image projection easier to understand and allows you to choose where you want to reduce the 3D effect without having to be someone who is an expert on panoramic image projection math. The textures are actually quite forgiving overall so they can be made with quite a large radius brush in your paint package.

* * *

Also, as a little way to say thanks for your patience, I've written a small javascript called "PanoView.ms" that can be used from inside the Maxwell Render GUI's script window to send your active latlong / fulldome imagery from the Maxwell render window directly to an Oculus Rift HMD or desktop viewer tool. Currently 9 of the most popular media player tools are supported in PanoView. You can read the script's source code for installation details and on what changes can be made to the tool's configuration.

Note that only some of the viewing tools can handle EXR files natively but every panoramic 360 degree viewer tool can show a PNG image.

Download the PanoView for Maxwell Render Script:
http://www.andrewhazelden.com/projects/ ... render.zip


Here is a full list of the viewing tools supported in PanoView version 1:

Kolor Eyes Desktop (Free) - LatLong Capable - Is the default viewing tool in PanoView
http://www.kolor.com/kolor-eyes/download/

Live View Rift (Free) - Fulldome and LatLong Capable
http://soft.viarum.com/liveviewrift/

Amateras Dome Player (Free) - Fulldome and LatLong Capable
http://www.orihalcon.co.jp/amaterasplayer/en/

Shotgun/Tweak RV Player - LatLong Capable - Has EXR Support
http://www.tweaksoftware.com/products/rv

Assimilate Scratch Play - LatLong Capable - Has EXR Support
http://www.assimilateinc.com/products/scratch-play/

Whirligig Viewer (Free) - Fulldome and LatLong Capable
http://www.whirligig.xyz/

Adobe Speedgrade - Has EXR Support
https://www.adobe.com/products/speedgrade.html

VLC Media Player (Free)
https://www.videolan.org/vlc/

DJV Viewer (Free) - Has EXR Support
http://djv.sourceforge.net/


Regards,

Andrew Hazelden
Domemaster3D Co-Developer

[Nova66]

Hi this is Andrew Hazelden.
I'm one of the co-developers of the Domemaster3D shader.
Today I created a Maxwell Studio .mxs version of the standard Maya/Domemaster3D mental ray scene called "LatLong_Stereo_CubeX".

Hi Andrew,

Thanks for your detailed explanation of the Domemaster3D shader settings, it's really helped to clear up a few things. I'm quite excited about getting the Separation Map working in the next Maxwell update

The CubeX sample scene is great because it'll make it so much easier to compare apples to apples. The camera placement in your anaglyph test render is a little unfortunate because there is no detail in the zenith & nadir regions to show up any warping at the poles. To be able to see any potential warping I've made some minor changes to the MXS file. Firstly by moving the camera just a bit so that it aligns with the origin (0,0,0), and secondly by adding two cubes at the zenith and two at the nadir so there is something to see when the warping is present.

Here's my test render of the scene and I'm hoping that when the next Maxwell update comes out, those red and cyan misalignments at the poles will go away:



Thanks for following up on this issue,
Andrew Novinc.

[seghier]

Thanks Andrew Hazelden for the informations .
Andrew Novinc ( Andrew and Andrew ) : are you try multilght with panorama ? or it's impossible ?