This is the Octane material test scene ... The irony here is that the model is build in MoI, it's not normal (sorry :P) to have a so bad dancing reflection result on this 33.000 triangles spheres.

Ewww. What do the devs say over in the octane forum? I confess to forming a certain.. opinion.. based on the earlier discusion that took place here, but I assumed progress was being made over in the beta forums. Are they saying this is working as intended or something?

Yeah, I'm still seeing artifacts as well despite the claimed efforts to correct them.

I agree with Michael, I don't believe smoothing groups is the answer...

-Will

"I'm still seeing artifacts as well despite the claimed efforts to correct them"

Same here. I just purchased Octane assuming this was fixed after Michael's lengthy dialog with Octane.

I know my MoI model is good because it renders smooth in KeyShot.

Ed

Hey guys, put some pressure on the MoI to Octane thread then,

http://www.refractivesoftware.com/forum/viewtopic.php?f=23&t=743&sid=92ebc45641680a2a39833b5cf8c9f914&p=13425#p13425

I can't have a discussion with Radiance myself anymore, looks I'm a bad little duck, never happy, and without any respect for the work done.

"You are not authorised to read this forum."

I'm not yet a customer because before I need that the problem with Moi will be addressed.
All the world of render engine (in particular GPU based) seem overpressure.
For now I pass, waiting to see what happen between arion, octane, shot, iray ....

But it's reality that all the application on the market nowadays are full of bug (except moi), missing feature and workflow issue. It's not only octane, that at least is still a beta.....

Hi Falcon,

Yop, I don't blame Octane, at all. I can understand bugs, beta etc, no problem.
The problem here, it's to make them admit there is a bug, I've never seen something like that before.

Last note from Radiance :

"the only remainins 'shading artifacts' is the intersections on low poly geometry with direct sunlight,
which is'nt an artifact, it's just something that's natural, there is no physically correct solution to it."

Have you seen the reflection sphere above ? :P ... seriously ... I don't get it.

The day he will consider that indeed, there is something stange, we all win. Then if it takes 6 months to fix it, I don't care.
The main problem is that he has no clue of what this sphere should look I guess.

"Last note from Radiance :

"the only remainins 'shading artifacts' is the intersections on low poly geometry with direct sunlight,
which is'nt an artifact, it's just something that's natural, there is no physically correct solution to it." "

I don't read the whole thread, but if I understand right ... so far I know it's a known general render technic problem. I can produce this kind of error per Vray too.

Make me wondering, maybe unbiased renderer can't deal with this 'special' normals surfaces ?

Will try to see if maxwell or fry can render them as expected ...

Hey Paq, can you post a picture how that sphere & reflections should look like if done correctly? I have a license for Octane and I'm willing to voice myself on the forums, but I don't know what I'm arguing about exactly :-)

Candide, check out the reflection of the cylindrical base at the bottom of the sphere.

I don't have access to the octane thread since I'm not a customer, but I definitely got the impression that Radiance had some walls up when presented with this issue in the earlier thread. I'm really disappointed to hear that they shut you out over there, Paq. You're one of the most respected members here and over in the modo forums, and I've never seen anything but helpful commentary from you.

->I can't have a discussion with Radiance myself anymore, looks I'm a bad little duck, never happy, and without any respect for the work done.

I don't believe that's true, your input and knowledge is most appreciated.

I don't know much about polygon stuff, I just know something doesn't work!

Marc

There's also a simple test file available here: CylinderOctaneTest.zip

That's a cylinder surface that is split into some slices. But the vertex normals on it come from the true cylinder NURBS surface, so creases or seams between the slices should not be apparent if those good normals are being used.

For example in Cinema4D importing the OBJ file from the zip file linked to above renders like this:



- Michael

Hi Michael

Just tried it both are faceted in Octane. For my understanding are the vertex normals in the .obj or in the .mtl file?

I hope this can be solved by the Octane crew. Thanks Michael for spending time on this as its not really your problem.

regards

Jamie

Hi Jamie, thanks for testing it.

> For my understanding are the vertex normals in
> the .obj or in the .mtl file?

They're in the .obj file - you can actually see them listed if you open the .obj file up in a text editor (try a text editor a bit more advanced than notepad, like notepad2 or notepad++), they're the "vn" entries that look like this:

vn 0.08659345 0.01440297 -0.99613961
vn 0.16999357 0.05883424 -0.98368731
vn 0.08664738 0.03098550 -0.99575706


When MoI exports an OBJ file, it puts the vertex normals from the smooth NURBS surface in there at the corners of each polygon that was created, so that when the polygon is rendered it can be shaded to closely resemble the original surface.

But if the stored vertex normals are not used and instead are just calculated by averaging the normals of adjacent polygon face normals together, it won't produce the same result.

- Michael

Hi Michael

The attached cylinder from blender works fine in octane. Any ideas why. I noticed that the vn entries in the blender file are shorter six figures after the decimal MoIs have eight. I have no idea if and why that would make a difference.

It does seem that MoIs files are different somehow. They will look at it in the next beta but if you have any ideas that would be great.

regard

Jamie

Michael's cylinder is designed to test that the vertex normals are superseding any shading groups / face normal averaging. The side of his cylinder is divided into 20 slices (polygon islands), each with their own shading group. If the vertex normals were being respected, it wouldn't look faceted between the slices. It works fine in Max, Maya, C4d, etc. Even in the opengl/directx viewports.

Hi Jamie - sorry maybe I wasn't clear with that cylinder test - the object in that previous cylinder test was not a "default cylinder" created by MoI, it was a test file set up similar to a model made up of multiple surfaces.

I made it by cutting a default cylinder into several different segments.

It's just an easy way to see what will happen on a more complex model made up of multiple surfaces that are joined together instead of just one big surface.


Probably your Blender one is just a default cylinder created in Blender? It most likely has not been cut up into different pieces, that's why there is a difference.


But there is also a pretty big difference between a polygon model that is created in a poly modeling program like Blender versus one that is generated from NURBS surfaces in a CAD program like MoI.


In a poly modeler you never actually have a "true" cylinder shape at all, it's all set up to work with polygons only and may use things like smoothing groups to separate polygons into different clusters that should have a kind of "fake smoothing" applied to them.


In a CAD program like MoI on the other hand, when you create a cylinder it is an actual cylinder spline surface. In a CAD program there isn't any "smoothing group" mechanism being used, things will look smooth if the actual NURBS surfaces are smooth, and creased if the NURBS surfaces are creased.

In that cylinder test file, the different NURBS surfaces are completely smooth to one another, so they are supposed to look smooth when rendered as well, despite them being in separate pieces. That's because the vertex normals in the file are supposed to be used for shading rather than just cooking up normals from the polygon faces.


> It does seem that MoIs files are different somehow.

It's not just MoI - it's stuff that is generated from any CAD program rather than a polygon modeling program.

CAD programs expect for a rendering program to use the vertex normals in the file to shade the objects, and not smoothing groups - smoothing groups should only be used for stuff coming from a poly modeling program and not a CAD program.


Note that in that same .zip file I also had an OBJ file saved from Rhino in addition to MoI - I thought you said that the Rhino file also showed the same segmented appearance problem in Octane as well?


EDIT: vodkamartini has described it more simply above as well.


- Michael

Hi Jamie,

> I noticed that the vn entries in the blender file are
> shorter six figures after the decimal MoIs have eight.
> I have no idea if and why that would make a difference.

No, that shouldn't make any perceptible difference.

The difference between those files has more to do with the way faces are organized with points shared between them.

The MoI example one is made up of several separate mesh pieces that physically touch each other but don't have "welded" points in common.

But they do have the same vertex normals where they are physically touching, that's why they should appear to be smooth with one another despite being in separate chunks.

If they don't appear to be smooth with one another, then it means that the stored vertex normals are not being used for the shading.

Every other renderer that reads vertex normals from the OBJ file will render that segmented cylinder as a totally smooth object without seams between the segments, but something seems to be different in how Octane is handling vertex normals that makes it appear different in Octane from other renderers.

- Michael

Hi Jamie, also one thing to note is that the default Blender OBJ importer also has a bug in it that discards the vertex normals instead of using them, so if you use Blender to test OBJ files with it could be easy to get a mistaken impression of what they are actually supposed to look like.

I have an updated version of the Blender OBJ import script which does not throw out the vertex normals here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=3164.11

Maybe this is a source of some confusion, if imports into Blender were being used as some kind of model of how things are supposed to work...

- Michael