lots of metal pipes and beams (hundreds)

Hello,

I need to make a virtual tour of an industrial building and I’m struggling to determine the best solution to render lots of aeration/liquid pipes, and beams to support all the metal floors. Also lots of hanging neon lights and railings. Right now all the lighting is baked, but unwrapping everything for lightmapping is a nightmare there are thousands small unwrapped faces, the baking process takes hours due to the large amount of lights/objects…
Is there a better way to do it ? Knowing every pipe and beam is slightly different from the other so instancing isn’t really an option…

Thanks in advance!

The best suggestion would be to take the environment into a 3D application, run an automatic unwrap on it with the least amount of distortion. Sew up the UV islands that you can, and basically make sure the lightmap UVs are laid out in the most optimal way they can be (i.e. minimal amount of islands, plenty of pixel gap between islands, and minimal amount of actual maps). This will also help speed up bake times as well. Downside is, you will have to do some manual work setting up those lightmap UVs.

Thanks for the reply!
Well that’s mostly what I do for now, flatten mapping in 3ds, but many uv islands are so small they end up covering 1 or 2 pixels in the lightmap texture.

Yeah that’s where some of the manual stitching work comes into play. Automatic will only take you so far, and there’s usually always a bit of manual work. It’s worth it though, as you get a better quality set of lightmap UVs.

Ok I understand!

Now I wonder if it’s better to have all the pipes in the same imported mesh, or detach all the pipes in separate mesh to be able to use occlusion culling on them? They are almost everywhere so the camera is always looking at it and the big mesh is always displayed.

Also I’ve been playing with LODs and wonder if they are detached, would it be a good idea to apply LOD group on each one to display a degraded version or cull them at a certain distance, even if the camera is looking in its direction?

I don’t know if LOD groups require a big amount of CPU or not and if it would actually be better or worse to have hundreds of LOD groups…

Sure take the lod groups some cpu , but are you limited by the cpu? If not it will probably help you to render less stuff then.

I think the CPU shouldn’t be a problem. In the end I’m using layerculldistances for objects with no smaller LODs.