Photogrammetry with specular/roughness Data Advice

Consue

Hi Im considering doing some photogrametry in a studio setting for some of my rocks and other assets and Im looking for how to go about this.
Is there a way to extract specular/roughness information from an object? If so how would I go about doing it?
Some of my rocks are less than 5 cm wide, is it worth it to scan them anyways?
I have access to studio lights (strobes) a turn table as well as high quality prime and macro lenses.
I also plan to process my scans with Substance Maya and Zbrush.

EarthQuake

General speaking the big boys (Quixel et al) do fancy stuff with polarizing filters and specialized light setups that will be difficult for you to replicate without a R&D and mechanical engineering budget.

But something simple you can try is getting a circular polarizing filter for your lens and twisting it to remove as much of the reflections as you can. This should get you a pretty good base for the albedo. You may be able to take two shots with the CPL in different positions, and... do some fancy math with that or something, to derive other maps.

Paul Debevec has done a lot of research on light stages / scanning rigs: https://www.pauldebevec.com/  - His work is probably beyond the scope of what you're trying to accomplish though.

@radiancef0rge put together a rig to do some of this a while back: http://chrisperrella.com/site/materialscans.htm - you could try asking him for advice.

@AlecMoody scanned an entire freaking car bit by bit - https://www.artstation.com/alecmoody - https://polycount.com/discussion/184167/development-log-building-a-complete-car-from-scans#latest - so he may have some input. I know he had some interesting techniques for scanning metallic objects (which are notoriously difficult).

Honestly though, you're probably better off using the scan for normals/displacement and maybe the base for albedo and hand tuning the other maps. For anything that isn't metal, specular is easy, it should be 4% / metalness = 0. Roughness can be eyeballed for the base values and traditional texturing techniques can be used to fill out the detail.

m4dcow

You would want to look into cross polarized lighting. I haven't done much photogrammetry myself but I have been messing with a material scanner setup for years, and if you take a cross polarized shot and subtract it from a a shot that isn't cross polarized (rotate the circular polarizer 90 degrees) with a bit of processing you should be able to derive a specular/gloss component. I'm not sure how you would end up matches the 2 sets of photos projected onto the mesh though but if you did figure out how to do that, cross polarized is the way to go.

The trick is being able to get these 2 shots without moving the camera. I have gone through many iterations myself, back when I had few lights I would rotate the filter on the lights, then I mounted a follow focus adapter to the polarizer itself which gave enough leverage to rotate the filter without moving the camera. http://rtgfx.com/pbr-texture-scanner/ this guy used a Fast Polarization Modulator, but he also talks about changing polarization at the photosite which is what Debevec does, which is at a given light position having 2 lights close together with polarizer filters rotated 90 degrees relative to each other, so when you fire one set of lights you get cross polarized, and then the other you don't.

My latest iteration has a stepper motor mounted on my camera that rotates the polarizer filter between shots, which is an idea I abandoned but later got it dialed in pretty well, you can check out all of that stuff I have done in my sig.

Even if you don't do any of the filter rotation stuff, cross polarized lighting gives you the flattest lighting which lends to good photogrammetry results, so while getting cross polarized lighting in the field might be tough with the sun, you can use it in a studio setting to great effect.

gnoop

Photogrammetry works not very well for reflective/ polished/ wet  things anyway.   So you would want to dust things over before doing capture.  So roughness is something to fake later really. Usually not so big a problem using depth and color texture.