Isotropix Forums

Interior GI test.

Clarisse related tips and tricks

Re: Interior GI test.

Unread postby armen » Fri Aug 09, 2013 8:16 pm

isoyann wrote:
"One of theses feature is the ability to texture the ray color exit of the GI."

Is that some type of light map? Final gather map?


Not really.. using distance clipping, the GI light will only shoots rays in a max range of 1 meter for example (if you set the clipping to 1m). Above this distance, the rays hit nothing and return black. The ray color exit allows you to replace the black with any color, or texture it. Using a panoramic 360 small render (32x32 is enough) of your scene to texture it (in clarisse you can use any interactive render as a map) give you a very good approximation of the full GI render.
If you change the scene (move objects, add or change lights..) , the panoramic render will be re-rendered and used on the fly , here is an example which show this with a very saturated colored lighting, computed in 20 minutes (there is still noise in the 1 meter GI range because the sampling is too low, but the walls, illuminated mainly by the ray exit color, is clean.. ).

Image
bedroom_clarisse1200sec par isoyann, sur Flickr


Isoyann, would you kindly tell a bit more of this setup?

For example:

1) Where did you place your panoramic camera?
2) I guess for the 32x32 render you created an additional lightset mimicking your scene setup but with real GI light?
3) How did you use the 32x32 image, specifically how did you map it to the light to make the lookup precicely match the panoramic render? In short: how did you make your GI light sample the 32x32 image correctly.

I'm very curious of this setup, because it looks like that, if set up correctly. you could use this approach to fake GI for distant or environment objects while using separate lightset and GI light for hero characters.

That would be so much better than my current workflow in which I use multiple point/area lights with their shadows disabled to fake diffuse GI irradiance to cut render time from hi-res animations to avoid using GI at all for environment objects.

If you don't have time to explain all of this, I'd be grateful if you could provide a scene file with this setup. I'm sure I'll figure it out =)
armen
 
Posts: 61
Joined: Fri Jan 25, 2013 11:33 pm

Re: Interior GI test.

Unread postby isoyann » Mon Aug 12, 2013 10:44 am

1) Where did you place your panoramic camera?


The panoramic camera was placed exactly at the center of the room.. the trick is to use a near clip for the 3d layer of the panoramic render, so objects very close to the camera don't mask everything.

2) I guess for the 32x32 render you created an additional lightset mimicking your scene setup but with real GI light?


The lightset was instanciated from the main render, so everything is automatically synchronized. Only the GI light is a custom one, to avoid an infinite loop (a light can't use a texture affected by itself..)

3) How did you use the 32x32 image, specifically how did you map it to the light to make the lookup precicely match the panoramic render? In short: how did you make your GI light sample the 32x32 image correctly.


The 32x32 image is used to texture the ray exit color of the main render's GI light, with the attribute "Sample exit color Texture" checked. About the texture attribute, it use a map texture (pointing to the 32x32 image), set to parametric projection, interpolation mode to mipmapping (very important to avoid noise), filtering mode to trilinear. Nothing is changed in the coordinates, by default the parametric coordinates give the correct oritentation for this kind of use, you don't have to change anything.
User avatar
isoyann
 
Posts: 640
Joined: Mon Jan 28, 2013 11:11 am

Re: Interior GI test.

Unread postby armen » Mon Aug 12, 2013 1:52 pm

Thank you for the answer.

There's something wrong with this method.

Essentially what we're doing is simulating Irradiance Cache map by creating panoramic camera and mapping it as a texture to GI light's exit color. I decided to test whether panoramic mapping was the right choice by creating an ambient light and setting its gradient such a way as to colorize the floor, walls and the ceiling:

Image -> Image

When I switched to parametric light I got this, which seems as if my map has been flipped vertically, and the lighting looks wrong:

Image

Whereas by setting map projection to Camera, through my panoramic camera, I got this:

Image

Which seems legit, but there are lots of spots. It's like in regular Irradiance Cache where there aren't enough samples to produce a clean result.

Switching to normal exit color texture confirms this:

Parametric:

Image

Panoramic:

Image

Would you please share scene setup you used to produce your image?

Here's mine: project_file

There is also a bug where if you try moving light in one of the main scenes with exit-color mapped GI light back and forth very soon Clarisse freezes and crashes, I'll report it in the bug section.
armen
 
Posts: 61
Joined: Fri Jan 25, 2013 11:33 pm

Re: Interior GI test.

Unread postby isoyann » Mon Aug 12, 2013 6:57 pm

When I switched to parametric light I got this, which seems as if my map has been flipped vertically, and the lighting looks wrong:


The lighting is right here.. don't forget it simulate bounced light.. so it is perfectly normal that the celing receive the red ground color, and the ground, the blue of the ceiling... You can test the setup in exterior with a big sphere with a gradiant.. if you use this to generate the panoramic and light a sphere with it, the top of the sphere will receive the blue of the sky, and the bottom, the brown from the bottom of the gradiant..
User avatar
isoyann
 
Posts: 640
Joined: Mon Jan 28, 2013 11:11 am

Re: Interior GI test.

Unread postby armen » Mon Aug 12, 2013 7:58 pm

Yeah... I'm sorry, I don't think so. Tried various setups and parametric doesn't look right at all.

But this doesn't matter anyway, because I found out that as soon as I activate "Enable distance clipping", global illumination luminance and color in areas that actually are still calculated get so noisy, GI becomes practically unusable it doesn't even matter which sampling values you throw at it. If sampling gets too high, render times skyrocket.

This is unfortunate, since the setting could be a nice way to cut render time.

Back to direct lighting only setups, there's no magic here either.
armen
 
Posts: 61
Joined: Fri Jan 25, 2013 11:33 pm

Re: Interior GI test.

Unread postby dea » Thu Jan 16, 2014 6:32 am

Hi All,

My name is Dea and I'm new here....just download the PLE and can't wait to try it.
I hope someone can help me regarding with render time & render setting.
I'm curious with render time, I saw from youtube that Clarisse can render billion poly
in couple minutes ( canyon scene ) - and how come the render time become 20 - 40 min
for one small bedroom with no texture on it ?
My big expectation is Clarisse could render interior and arch render fast...and I can adapt
some for my team.

Many thanks,
Dea
dea
 
Posts: 2
Joined: Mon Jan 13, 2014 5:32 pm

Re: Interior GI test.

Unread postby sam » Thu Jan 16, 2014 10:21 am

Hi Dea,

Welcome on board!
Actually this is a very interesting question. Unlike rasterization (Open GL for example), render times with raytracing don't increase linearly with the number of polygons.

The rasterization method basically draw triangles. Let's say it takes 1 second to draw a triangle then it will take 10 seconds to draw 10 triangles. This is a bit over simplified but you get the idea. The more triangles there are, the longer it takes to draw those triangles.

Using raytracing, you don't draw actual triangles. You launch rays and find which primitives they intersect. Naively, someone would try to intersect every single primitives with the ray until the nearest one is found. A smarter approach is to organize spatially primitives so you can discard quickly most primitives. In the real world, you end up intersecting a very small number of primitives to find which one is the nearest one. I invite you to read http://en.wikipedia.org/wiki/Quadtree if you want more info about that magic ;)

Now back to your question. Why is it slower to render an interior (with a rather small number of polys) than the canyon scene (with insane amount of polys)? As I wrote, it's not a matter of polygons but it's a matter of number of samples required to find where the light comes from.

In the canyon scene, 90% of the lighting comes from a direct light (a sun). The rest comes from indirect lighting computed using global illumination algorithm. When the lighting is direct, it's really easy to the renderer to know where it comes from. With a few number of samples you get very good results. When it's indirect, the renderer has to shoot blindly many rays to resolve indirect lighting (light bouncing all over the place). The less samples, the noisier the image gets. The more samples the better the image gets.

In an interior scene, 90% of the lighting is indirect. A rather small amount of direct light goes through the window. Unfortunately, in that configuration, you need to increase greatly the number of samples of the global illumination.

Hoping it clears things up,
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 1559
Joined: Fri Jan 25, 2013 11:33 pm

Re: Interior GI test.

Unread postby dea » Sat Jan 18, 2014 10:34 am

Hi Sam,

Thanks for your explanation...it helps me to understand the difference between exterior and interior scene.
And from this point, I think Clarisse can help me in exterior arch viz only, not in interior viz scene, correct me if I'm wrong... ;)
I have big expectation on Clarisse and hope that Clarisse can be robust tool to do both scene. If Isotropic could give some
complete interior images created by Clarisse with rendering time, and some trick and tips to create interior scene with
robust interactive.

Many Thanks,
Dea
dea
 
Posts: 2
Joined: Mon Jan 13, 2014 5:32 pm

Re: Interior GI test.

Unread postby sam » Sun Jan 19, 2014 10:36 am

dea wrote:And from this point, I think Clarisse can help me in exterior arch viz only, not in interior viz scene, correct me if I'm wrong... ;)

Actually I just meant, for an exterior scene lighting is straight forward Sun + GI. For interiors, if you don't want high render times to get rid of the noise, you need to help the renderer by adding some direct lights.

I think Yann may give you a hand on this.
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 1559
Joined: Fri Jan 25, 2013 11:33 pm

Re: Interior GI test.

Unread postby Arvid » Mon Jan 20, 2014 12:17 pm

Considering how the Clarisse render engine operates, does it allow for something like a portal light, which concentrates rays and samples to a smaller area such as a window? In mental ray, it's a great way to light interiors using exterior diffused light. :)
Arvid
 
Posts: 191
Joined: Thu Jan 31, 2013 3:44 pm

PreviousNext

Return to Tips and Tricks