Procedural Rocks vs Photogrammetry Rocks

[MagnusL3D]

I've search around alot and read several threads on procedural rocks, both here on odforce.net and sidefx forums and elsewhere. And i've tried alot myself, but when trying to create procedural rocks in Houdini I always (and others) always seem to fall back on the worley noise and you get a rather distinct pattern with that, and I feel it's mostly suited for rocks up to a certain size. 

 

Some of my attempts:

 

 

Is there any way to get a more organic flow on bigger elements, rocks walls etc, like the Phorogrammetry Rocks from the newly relased game Vanishing of Ethan Carter. 

Something like this ?

 

https://p3d.in/1tNIW

 

It doesnt have to be 100% similar but, I am thinking more of the shapes beeing more sudden and distinct, and not to predictable.

Edited September 28, 2014 by MagnusL3D

[Guest tar]

Never seen a generated noise that creates fracturing and erosion and dirt that you see in a cliff face.

[eetu]

Don't forget Macha's awesome rocks, http://forums.odforce.net/topic/9532-rock-hard/

[Guest tar]

Macha's rocks and some dirt filling up holes/on ledges may work well

[MagnusL3D]

I've seen Macha's rocks several times, and even though they  are awesome they are still single stones which (to me atleast) is alot easier to make look unique. 

 

I am not sure what I am looking for was just throwing the question out there to see if something interesting came up.

[eetu]

I very much know what you mean, I've tried it a few times over the years with not much success. As Marty implied, a completely local procedural/fractal solution will never really cut it.

What I would do next would probably be; 1) study some more geology, crystal structure and preferred fault lines in different materials etc 2) Try and create geometry with that in mind, break things up 3) move to vdb sdf, and do the erosion and the like there. probably back to polygons again. 4) Do the dirt deposition in a shader

Of course very much depending on the type of rock..

[freaq]

you can combine both approaches...

use photogrammetry rocks, make them into VDB's and "composite" them into new or generic shapes.

or use a smooth and use a difference to derive at a displacement mat incl texture etc. so you can sculpt any shape and the "texture"
becomes a 3D rockifying material.

 

the two methods need not be mutually exclusive.

also photogrammetry is not perfect:

lighting is baked in,
tones of little holes that need patching
can only create that specific scanned rock, not arbitrary shapes.

hard to art direct.

need to go to a site to get data. etc.

Edited October 6, 2014 by freaq

[eetu]

also photogrammetry is not perfect:

lighting is baked in,

 

If you've acquired the data in ambient lighting conditions (as you should), you can do an ambient bake in 3d without the textures but with the same uv mapping, and then divide by that to remove much of the lighting from the photogrammetry textures..

[Guest tar]

There are tools like Bitmap2Material that try to produce textures

 

http://www.allegorithmic.com/products/bitmap2material

[freaq]

If you've acquired the data in ambient lighting conditions (as you should), you can do an ambient bake in 3d without the textures but with the same uv mapping, and then divide by that to remove much of the lighting from the photogrammetry textures..

granted, but it is a constraint one has to work in again, but I digress

 

I agree photogrammetry is a very useful tool, but proceduralism and photogrammetry need not be competing,

and can often be even more powerful when combined

[MagnusL3D]

I agree that photogrammetry and procedural generation do not need to compete.

 

The goal for me would be to be able to create mile long cave systems that have interesting organic rock-like sharp features, if they are based on photogrammetry elements or noise or a mixture is fine.

 

My current solution is to build a lowpoly mesh "hand made", up-res it with the "remesh" SOP, innhert the UV's from the hand made lowpoly, have a Z-brush displacment map, and make a falloff attribute for the UV-seams, and make the texture displacment falloff close to UV-seams and replace it with Noise along the UV seams.

 

This approach work very well, because I can then use a Diffuse and Normalmap that "syncs" up very nicely with the displacment map. 

 

The bad part with that solution is the ridiculous amount of polygons it generates and that the polyreduce SOP is single threaded and destroys the UV's, so I have to "re project them" from the original mesh with a attribute transfer SOP which does it works sorta of semi bad because of the big polygons going over the uv-seams in the original mesh. I tried the UV-Relax SOP on Orbolt that solve that issue to a degree, but messes up some other parts of the UV even more terrible, and for some reason I cant put that SOP in a Foreach. 

 

But in the end, very close to a good result but not crossed the finishline.

[edward]

I tried the UV-Relax SOP on Orbolt that solve that issue to a degree, but messes up some other parts of the UV even more terrible, and for some reason I cant put that SOP in a Foreach. 

 

You're probably running into the problem where the inside the HDA, there's already a ForEach SOP that's using the default stamp name. So if you put that in another ForEach, try changing the stamp name to something else.

[MagnusL3D]

You're probably running into the problem where the inside the HDA, there's already a ForEach SOP that's using the default stamp name. So if you put that in another ForEach, try changing the stamp name to something else.

 

Aha ! I will try that, thanks alot !

[el_diablo]

Maybe you could find a way to do 'texture bombing' with procedural rock textures at various scales and distributions. I have seen some nice results from it that hide the noise basis for plethora of uses.