I can’t get Occlusion Culling to work at all. I’ve followed the steps in the documentation several times, but the camera is only doing frustrum culling. Here’s a screenshot from a very simple test scene, with default settings in the Occ.Culling window, all objects marked as static, baked. Is there anything obvious I’m missing here? Any camera settings, anything?
according the screenshot you didn’t calculate any occlusion culling cause I don’t see the view cell hierarchy at all there.
a screenshot that shows the OC settings would potentially explain more.
also, the angle of your view does not say if it is really occluding the whole view cone, otherwise it might just happen that it shows up as a single piece of the view bounds of the wall behind not being occluded will ensure that it is completely rendered.
I have not had any success in building occlusion, even with a simple scene. We set everything static and the Umbra program crashes after it runs for a while. The progress bar never moves much. We did have success with ealier beta versions.
I think I’ve figured it out–the OC cells are just tall enough to peek over the nearby objects, so even though the distant objects are very much out of the camera view, they still get drawn due to the particular cell sizes. Is there more detailed documentation of the OC settings? How can I make the cells shorter (shorter than my buildings, basically)? I’ve got the cell size set at .01 but it’s still too tall. What unit is that? Is there a big downside to making them .001, or can I do something else to shrink them?
I had my first try with Umbra and so far it completely sucked.
a) The documentation is quite bad on how you setup a working scene that gets culled properly.
b) What you easily can watch is frustrum culling but occlusion culling is very picky to setup and i’m still not sure what you have exactly to do to get it always working.
c) Calculation times are insane for very simple setups and sometimes the machine hangs. Cancel doesn’t work and you have to kill the task the old skool way.
If you’re setting up a very simple scene, this realtime 3d package more turns more into some sort of rendering software, although i get decent pictures out of Modo in less time. Maybe i’m remembering things wrong but in early 3d days occlusion culling, maybe on a much more simple basis, was done in realtime.
Anyway is there any reasonable help which explains how to setup things correctly and why it works this way? Trial and error is the wrong strategy for these calculation times and the time i can spend on it.
Secondly and although due to the calculation times it’s questionable but nevertheless can you also use Umbra on procedural during runtime generated objects?
Thanks
I don’t understand why it’s a problem setting up Umbra so that it works 100% in such simple scenarios
@taumel the character controller falls into the abyss … a little collision please ?
I have serious problems in a big scene Umbra crashes a lot …
a
./
Sorry, i had no time to look further into this but i’ve updated the ambient demo with a now working occlusion culling.
Btw is there any reason why…
a) a translation function of a rigidbody is named different than a translation function of a transform and
b) why there is no support of local and global spaces with a translation of rigidbodies?
If not, then it would be nice if this could be supported.
Anyway back to Umbra, i still don’t know why it needs a few tries until one really works but i guess that there are just a few bugs left, so time will tell. Btw this one is larger due to that i wanted soft lightmaps without compression artifacts, and as most of the time with browser content, best watched in fullscreen mode. Btw i’ve never experienced the camera falling through the ground, is this still the case?