instance problem?

Hi!

I am having some strange errors with a huge radiance scene here. As you can see, some surfaces are rendered as black (the region over the arches). These errors arise since I added more and more geometry. The model is made up of a lot (!!!) of instances, to get all the geometry into memory, but I know that there is some problem with overlapping instances. So is this a symptom for such problems, or am I experiencing something completely different here?

I attach the small pic, which should be ok for the list. Thank you for any help, CU Lars.

Hi Lars,

Can you offer us a clue as to what this image is _supposed_ to look like, or what version of Radiance you are using?

-Greg

···

From: Lars O. Grobe <[email protected]>
Date: April 16, 2005 7:52:46 AM PDT

Hi!

I am having some strange errors with a huge radiance scene here. As you can see, some surfaces are rendered as black (the region over the arches). These errors arise since I added more and more geometry. The model is made up of a lot (!!!) of instances, to get all the geometry into memory, but I know that there is some problem with overlapping instances. So is this a symptom for such problems, or am I experiencing something completely different here?

I attach the small pic, which should be ok for the list. Thank you for any help, CU Lars.

Can you offer us a clue as to what this image is _supposed_ to look like, or what version of Radiance you are using?

Hi Greg!

There is an image mapped onto the region over the arcs. I attach an older image where I had less geometry in the scene. Here the black region does not appear.

I am using radiance 3.6.1, compiled it on a x86 machine.

CU Lars.

Hi Lars,

I'm only guessing, but is it possible that you have two surfaces coincident in that plane? There's no way to debug this problem based on a pair of images. I haven't seen a rendering problem similar to this before.

-Greg

···

From: Lars O. Grobe <[email protected]>
Date: April 16, 2005 8:59:06 AM PDT

Can you offer us a clue as to what this image is _supposed_ to look like, or what version of Radiance you are using?

Hi Greg!

There is an image mapped onto the region over the arcs. I attach an older image where I had less geometry in the scene. Here the black region does not appear.

I am using radiance 3.6.1, compiled it on a x86 machine.

CU Lars.

Hi!

I'm only guessing, but is it possible that you have two surfaces coincident in that plane?

No, I don't think so. I checked everything this morning. The silly and interesting about this is that I get the wall behind the affected surfaces as some kind of mask. I attach one more rvu-screenshot to illustrate this. As you can see, the surface is black, if there is a wall behind, and bright (colorpict-modifier visible) if the window opening (or no geometry at all) is behind. So it seams to be some kind of geometry mess that I created here. Also, this behaviour changes with oconv -r settings.

I use instances heavily in my scene. I have been warned not to do so (and of course, most of my instances don't fit into a cube, but are all kinds of random geometries). On the other hand, the model is too huge to be rendered without the use of instances, and I also cannot divide the model into some cube-like parts, as there is a logical structure of building elements in this building which defines how the model is built. So maybe I just corrupted the scene by using too many (overlapping) instances' octrees? This is the only explanaition I could imagine at the moment.

TIA+CU Lars.