Brush stencil samples - 3D scan experiments

fisj

I’ve been working on a project to scan surfaces and make them into alphas and stencils for zbrush/mudbox. I've got a link to some samples below to try out. These surfaces are geared mainly towards high poly game asset creation.

Polycount is a pretty technical art community and you guys might be able to think of ways to use these that I haven't.

Its an eclectic mix of surfaces. Most examples are tiling, some also include the diffuse color of the surface. I’m mainly interested in ways to incorporate these into your existing workflows, what types of surfaces zb/mb users would like to use, and if there are any technical requests that would make it easier to use them.

Brief notes about the method:

The scanning method used is based off of “photometric stereo”, and gives a per pixel normal map of the object surface along with the flat color (diffuse) image map. The normal map is integrated into a heightfield and saved as the format of choice. The scan resolution is as large as the input photos. I mainly stick to bas relief style surfaces, since they’re easy to manage as textures and depthmaps. Tiling is done using texture synthesis to make them seamless, sometimes it works, sometimes the tiling is a glorious failure.

Anyway, try them out and let me know if you have feedback. I'd be happy to answer any questions.

Samples download page
Large image link [flickr]

gilesruscoe

Thanks so much, these are awesome

ENODMI

Snagged em, thanks for sharing. Ill be sure to give you some feedback once I try them out

AlecMoody

This looks really cool. I checked out your blog but didn't see any details about how you are creating these scans. Is it a photographic system?

fisj

Hey Alec, I haven't got a lot of technicals on the site, trying to put the emphasis on the textures themselves.

Yeap its photographic. The technique is called Photometric Stereo. The basic idea is you have a camera and a light, and you take 3 or more photos of the surface with the light in different positions. Using lambert's equation i = pN.L you can determine the normal of every pixel in the image.

If you take more than 3 photos the system becomes overdetermined, and you can start compensating for outliers like shadows and specularity.

The biggest strength of the method is that its per pixel. Got a 24MP camera? Then you have a scan with 24mil verts, the detail is great. The downside is that it breaks down for low frequency positional data. For example, scanning faces will result in some poor results for the overall shape. This is because the normal map generated from this method needs to be integrated into a height field. Since normal maps lack true depth, such as sharp discontinuities, low freq shapes get distorted.

BUT the skin pores, scars, hair will all be incredibly crisp. If you're painting depthmaps, that doesn't matter, and honestly bas relief data is easier to work with in mudbox and zbrush.

Also, in addition to a normal map, and depth, you get a flat shaded diffuse map. I know that artists painting face models love models with ambient lighting to reduce specularity and shading, and thats exactly what you get here.

-Paul

Shoy333

Thanks for these! Download and will be testing them shortly, Thanks again

sltrOlsson

That is very cool! I've always bean wondering what grass would look like if scaned, wink wink

I'm gonna try em out! Thx for sharing!

ceebee

Thanks mate! These are really cool, look forward to giving them a shot.

seth.

downloading as I type, thanks tons for these Paul, I will have a play with them.

Lennyagony

Downloaded and will give these a try.

Also wanted to add that this is an fantastic idea, looking forward to your first set/pack .

nyx702

Wow thanks for the freebees!
This project seems really awesome. I will be following this!

eyo

such a cool idea! thanks for posting it up, will definitely give these a try. As far as surfaces go, you can never have enough types of rock especially the non noisy variety.

Butthair

That's a cool method, it blows my mind and now I'll probably be on wikipedia for a couple hours reading about this stuff. Thanks for putting them up too, they look really nice, very crisp like your saying, but I have one problem...... Why matcap Red!? Why!?

fisj

Glad you guys like the idea, and the scans.

sltrOlsson, I'll try and get some grass scans this week, it will depend a lot on how windy it is.

If you guys keep watch of my twitter feed, you'll get updates when I post new samples. Thats probably a better solution than spamming forums every time I have something to show or say.

http://twitter.com/surfacemimic/

I have some real lizard skin scans that came out well, and I'm itching to try some alpha brushes with those.

@Butthair ... Matcap Red, because I'm a noob. Still not settled on the best material for these, I'd love to hear suggestions.

-Paul

looprix

T H A N K S!
Looking forward to see more from surface mimic

Daven

All of my love for you.

AlecMoody

I have been thinking about how I would integrate this sort of scan information into my workflow and I think for me the most useful samples are the lower contrast ones. The skin samples in particular. Granted, that's partially because I avoid adding strong micro texture to most of the stuff I bring through zbrush. Also, the tiling trim type textures like your stitches look like something I will definitely use.

Also, I would love to see some scans of mostly smooth concrete. That kind that has the sealant over top with super localized cracks and pores.

Edit:
Are you able to compute a specular component from the captures?

fisj

Hey Alec,

That makes sense. Skin and textiles are probably the most desired subject matter. On the zbrushcentral forums the tiling seam texture was pretty popular too. Textiles are pretty quick and easy to scan, clothes are plentiful.

Concrete ... I'm sure I could find appropriate concrete in NYC, its certainly plentiful. I'll bump it up in priority list. If you have an image example from google images, shoot me an email with that attached.

With regards to specular ... the short answer is no. There are similar photometric stereo methods that could do it, but lighting would need to be controlled a lot more precisely than I currently do, or would be able to.

Indoor scanning, such as face scanning is a different matter. Paul Debevec (the scanning godfather) has some amazing scanning rigs that he's built over the years, and he separates out specular and spec normal maps using polarization of the incoming light. Slick stuff.

Thanks for the feedback!

-Paul

fisj

I added four more samples for those requesting textile seams. Enjoy!

Textile Samples Page

- Paul

Envart

These are awesome! Thanks very much

fisj

Hey polycounters. I wanted to let you guys know that I've finally launched Surface Mimic after 2 months of hard work building more alphas (and setting up the company and 3d scanning process in general).

Thanks to the guys who tried out my earlier 3d alpha scans and especially those who gave feedback. It was invaluable.

This week I have four ZBrush and Mudbox alpha packs available on our site.

10 Leather Alphas
10 Torn Jeans Alphas
10 Hoodie seams and material Alphas
10 Tree Bark Alphas

Alphas are also available individually, if you'd like to try out just one or two. I hope you like them, and as always am grateful for feedback.



Full disclaimer. These alphas are for commercial purchase, but there are free samples available here. Sorry for the slight thread necromancy, but I thought posting in the same thread would be less polluting.

lloyd

could you go into more detail of how you made these, maybe photos of your setup so we could learn and try to make some ourselves?

seems quite amazing

fisj

Hey Lloyd. I've had a lot of people interested in the scanner tech specifically, so I wrote a page on the site explaining the process a little more here.

As far as making these yourself, the best example I've seen so far is probably Ryan's example of normal map photography. His method isn't strictly correct, but it covers the basic idea, and some decent results can be attained. The correct method is explained in detail more here.

Aside from the camera, tripod, light and calibration ball (an 8 ball on a stick to measure the light vector) all the heavy lifting is done in code. Because I use so many photos of the same surface, my application uses about 10 GB of ram for a 10 megapixel image. So, as you might imagine, its a pretty custom setup that isn't really suitable for public consumption.

Glad you like it.