bump mapping break-ups

Use this forum for general user support and related questions.
Post Reply
User avatar
lacilaci
Donor
Posts: 1642
Joined: Fri May 04, 2018 5:16 am

bump mapping break-ups

Post by lacilaci » Thu May 30, 2019 3:01 pm

Sometimes bump mapping works even if it's stronger, sometimes it starts doing weird things in different lighting setups...

Main point is that while bump mapping in luxcore can really look great, it is extremely easy to break the whole shading with it, even low values.

In this case, object edges are turning black. What's making this really bad, is that only extremely low value will keep edges intact, but the moment I use value that actually shows some bump mapping the moment I get problems and shader starts to look dirty on edges.

1mm - kinda safe value in this case, but under some circumstances I saw shader acting wrong, almost as if flipped normals
1mm.png
2mm - bump mapping starts to appear, but a little too weak, however already I have problem with the material
2mm.png
5mm - a bit too strong but in the ballpark of what I was going for, but it's just unusable without breaking the visual
5mm.png
What is weird, when I added additional lightsource (area light) one side stops showing problems, while area in shadows looks still wrong.
5mm_additional_light.png
This is just one example of issues with bump mapping, there is always another one that makes object look like if it has flipped normals or something when areas in shadow start to look bright and the ones in direct light become darker, this can happen even with very low bump map strenght... Will post some examples when I get the time and a good example case.

kintuX
Posts: 477
Joined: Wed Jan 10, 2018 2:37 am

Re: bump mapping break-ups

Post by kintuX » Thu May 30, 2019 4:26 pm

Found that Bump is broken using OCL.
Test, example, scene here: Re: Help to implement Bump mapping.

User avatar
lacilaci
Donor
Posts: 1642
Joined: Fri May 04, 2018 5:16 am

Re: bump mapping break-ups

Post by lacilaci » Thu May 30, 2019 4:45 pm

Hmm, I hope this can get solved. Today I did some furniture closeups and I was very happy with the result(shadow terminator aside) but then I moved camera further away and I was shocked.

Seems that several things contribute to weird looking shading when using bump maps.

UV mapping must be present or shading gets completely fucked.

UV seams can become very visible with stronger bumpmapping and it might even look like the mesh is bending near UV seam(I've seen this happening with a normal map only so far though)

Stronger uv mapping can make shadow terminator very visible

bump mapping value cannot be simply decided by how it looks under one lighting condition and camera cause a change in lighting might reveal some issues - this is super annoying.

I think the whole user experience here needs to be, higher value = stronger bumps and predictable stable results. Right now it is super time consuming to find out value that both looks good and doesn't break material, and sometimes it just breaks anyway...

I still say though ,when it works it looks much better than what I've seen in cycles for example so I hope this can be fixed without making it look... shitty.

User avatar
FarbigeWelt
Donor
Posts: 883
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: bump mapping break-ups

Post by FarbigeWelt » Fri May 31, 2019 12:37 am

For me your pictures show expected behavior. Bump leads to shadows. The higher the bump the darker parts of shadow. Depending on light source, direct, indirect, spot, area, angle bump maps throw different shadows. What happens if you throw a beam almost parallel along the floor. Yes, you can see every dust particle. Things made from lines have edges. This edges can throw ‚horrible‘ shadows like the long shadow tails of the bright dust particle. Somehow this effect is not avoidable. It might be smoothed using objects with higher resolution but never perfectly.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 883
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: bump mapping break-ups

Post by FarbigeWelt » Fri May 31, 2019 8:41 am

Looking for another topic I found randomly an article that should explain what I meant with my former post.
Just read the text below the images in this file.
3D-Printing Spatially Varying BRDFs ,
Olivier Rouiller ■ Technical University of Berlin Bernd Bickel ■ Disney Research Zurich
Jan Kautz ■ University College London Wojciech Matusik ■ MIT
Marc Alexa ■ Technical University of Berlin
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
B.Y.O.B.
Developer
Posts: 3009
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: bump mapping break-ups

Post by B.Y.O.B. » Fri May 31, 2019 9:33 am

It might be a similar problem like this in Cycles: https://developer.blender.org/D2574
But I'm not sure.
Support LuxCoreRender project with salts and bounties

User avatar
lacilaci
Donor
Posts: 1642
Joined: Fri May 04, 2018 5:16 am

Re: bump mapping break-ups

Post by lacilaci » Fri May 31, 2019 1:37 pm

here's how extreme bump mapping looks like in cycles. It is so strong that even cycles starts showing some shadowterminator issue but this is just to illustrate
cycles.JPG

User avatar
lacilaci
Donor
Posts: 1642
Joined: Fri May 04, 2018 5:16 am

Re: bump mapping break-ups

Post by lacilaci » Fri May 31, 2019 1:39 pm

I can of course break cycles too, but I do have to go to absurd values that would have been useles anyways and yet still the issue looks very different from what luxcore shows even with small bump mapping values.
cycleshigh.JPG

wasd
Posts: 250
Joined: Tue Apr 24, 2018 7:20 pm

Re: bump mapping break-ups

Post by wasd » Fri May 31, 2019 3:05 pm

From my observation, bump is broken in luxcore. Height map is just unusable because of unpredictable behavior. Normal map is weird at curved surfaces. Sometimes it looks like normals are inverted. In cycles there's no such problem.
When testing it's better to avoid random patterns, because it's hard to tell what to expect of bump (when it's random).
Image
bu.png
UPD:
Pictures above not very correct though. There's light reflected from the floor.
This one is without floor.
bu2.png
Left gap is weird, right?
Well, to this point, I'm not sure about my examples. But there's something strange with bump. And I'm yet to find what exactly.
I also found this https://blogs.unity3d.com/2017/10/02/mi ... h-tracing/
There's paper on microfacet bump and mitsuba code.
CPU Bidir (sometimes GPU Path) + Metropolis | Log Power | Core i5-4570

User avatar
lacilaci
Donor
Posts: 1642
Joined: Fri May 04, 2018 5:16 am

Re: bump mapping break-ups

Post by lacilaci » Fri May 31, 2019 4:21 pm

Yeah it's pretty much my experience all around ,have to be careful about maps, values, uv seams and yet it can still go wrong...

In one scene I had by accident changed specular value in shader and that stronger reflection on bump mapped surface crated a ton of noise that just wouldn't go away.

Post Reply