Load of old bowls!
I made this!
And by 'load', I mean one. And by 'old bowls', I mean a lovely piece of 16th century Spanish lustreware.
Zero lols, I'm afraid (unless it really is that bad) - it's a 3D model I've created (using photogrammetry) for part of my course, and as there are some talented 3D artists/modellers on this site, I'm shamelessly asking for feedback. Honest feedback, too - I'm well aware the model ain't even close to perfect...
[edit] Oh, and you'll need a webGL browser. Works best in Chrome, as usual [/edit]
Shit, my first link, shall I just bend over now?
(
FridgeBadger, Fri 8 Feb 2013, 11:37,
Share,
Reply)
I'm sure it's real lush but doesn't work on this IPad :-(
Enjoying your £9,000/year course fees?
(, Fri 8 Feb 2013, 11:47, Share, Reply)
EngD, so my course fees, and indeed living expenses, are paid for by your taxes :)
Sorry.
(
FridgeBadger, Fri 8 Feb 2013, 12:09,
Share,
Reply)
I absolutely love that
what equipment did you use? I'm considering buying a 3d scanning thingmybob for scanning antiques for reproduction in the near future
(, Fri 8 Feb 2013, 11:49, Share, Reply)
Two Nikon D700s bolted together and a lot of fancy software (and clever people - not me). I've done a bit of laser scanning, including with a 500k Arius colour jobby, but honestly, I think photogrammetry is the future - especially if you're after a nice attractive output rather than engineering-style accuracy.
You can try photogrammetry for free at sites like www.123dapp.com/catch
[edit]Btw, we also scanned the bowl with a Nikon portable laser scanner - to get geometry but no texture, but ended up not using any of the scan data (there was an issue with the scanner, but ultimately there was just no point, the photogrammetry got the geometry pretty well)[/edit]
(
FridgeBadger, Fri 8 Feb 2013, 12:11,
Share,
Reply)
Not bad but it loads upside down unless my vast knowledge of 16th century is wrong and it never is.
Never been a huge fan of voxels except in this case where you want to create an accurate record.
Solid works man myself and I would have made that out of a single revolve.
(
Cerebus Sticky! Sticky! Stick! Stick!, Fri 8 Feb 2013, 11:55,
Share,
Reply)
What's solid works? Does that mean you'd create a profile and make a solid object by spinning it round an axis? Unfortunately we were trying to create an accurate copy (we didn't manage it...), and the bowl isn't really very symmetrical...
And yes the upside down floral hat problem (see below) is do to with the viewer I'm using. When I've got time I want to transfer it to an open source version so I have more control over stuff like that...
(
FridgeBadger, Fri 8 Feb 2013, 12:14,
Share,
Reply)
I'd like to see an algorithm that converted voxels to CSG
I only use Povray myself.
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 12:22,
Share,
Reply)
Not sure, and getting out of my comfort zone, but I don't think it's using voxels. It's a point cloud, so the points are distributed arbitrarily rather than in a 3d grid... does that make sense?
Fairly simple to mesh a point cloud, could you then use that?
(
FridgeBadger, Fri 8 Feb 2013, 12:28,
Share,
Reply)
CSG isn't meshes,
it's things like spheres, cylinders and cubes, combined with union, intersection and difference. For example you can define a cube and take a spherical chunk out of it. Which tickles my maths gland.
*Constructive Solid Geometry
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 12:33,
Share,
Reply)
It's not that simple to mesh a point cloud properly.
you can get a reasonable first approximation, but there are cases where most of the algorithms break and manual intervention is required. Worth the effort if you want to render without holes or analyse the data though.
Adding normals and reflectance info to turn the points into surfels is another alternative that sometimes works well for rendering, unfortunately you still get problems with holes where the samples are too far apart.
If you want to go all the way to crazy you could just take hundreds of pictures and go for light field rendering. No geometry at all, just clever warp & blend operations.
(, Fri 8 Feb 2013, 12:51, Share, Reply)
True, though those algorithms are improving all the time - but yes, the mesh will usually need a lot of cleaning and work to make it properly solid, so it can be printed out, for example.
As for normals, hopefully you can see on this model that there are some! They were all calculated after capture though. There weren't many holes in the data, but huge amounts of noise thanks to the incredibly shiny material. Cleaning that up did introduce holes, which then required filling using the software's algorithms, which is why the model looks rough (in the actual sense of the word) in places.
We got this model using about 30 image stereo-pairs. My colleagues here have got this method (depending on the source object) to output geometry that's accurate to less than 20 microns, better than pretty much any laser scanner. We're looking at buying a light field camera to play with!
(
FridgeBadger, Fri 8 Feb 2013, 13:11,
Share,
Reply)
Spatial res on the light field cameras I have seen is unimpressive.
Probably best off sticking with a high res camera + robot type rig.
I wasn't getting decent lighting unfortunately - it was all flat. I believe it is there tho. Software at my end is the likely culprit.
Got any papers btw? I'm fairly interested in this stuff but haven't followed it for a couple of years.
(, Fri 8 Feb 2013, 13:24, Share, Reply)
It takes a while for the lighting to 'switch on' - was the model fully loaded? Lots of pink specularity...?
Funny you should say that about high-res... our first model used the full res of the image, and we got a very dense point cloud but with a lot of noise and holes. Running the same images but downsampled 2x, we got a sparser point cloud but cleaner data and fewer holes - probably because the feature detection worked better on lo-res images rather than high-res with everything 'smeared out' over more pixels.
And not sure what you mean by 'robot' rig - we just used a tripod and shifted the camera round! We did use a kinect to make a very rough 3d model first, and then used an algorithm in development here to predict the best camera positions for the object (though I doubt it would matter much with a fairly 2d symmetrical object like this). A couple of colleagues are actually building a robot rig which would take pics from all the best spots...
As for papers, I try not to get involved in the maths/technical side, but google scholar should throw up a load - try searching for just photogrammetry or 'structure from motion'. I get the gist of most of them, but they tend to lose me as soon as they start with the maths...
(
FridgeBadger, Tue 12 Feb 2013, 13:28,
Share,
Reply)
Povray is the CG equivalent of gimp
A complete crock, but popular coz it's free.
(, Fri 8 Feb 2013, 12:43, Share, Reply)
I only dabble so it's not worth paying for anything,
but I've seen some impressive images created in it.
Also I really like typing the co-ordinates in instead of having to faff around with the mouse and doing things by eye.
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 12:47,
Share,
Reply)
The CSG engine and scripting side is nice enough to use
It's the back end and implementation I've got no time for. CSG trees get unwieldy quite fast, and the use of implicit geometry for everything has been hobbling the performance for years.
By all means use CSG, but FFS expand the geometry, stick it into some *decent* spatial data structures and take a tenth of the time to render a scene!
[edit] They can't actually do this, because the CSG engine allows infinitely large objects. This means that to have something with decent performance for, say, triangle meshes and tessellated objects they need to support 2 data structures simultaneously and decide what can be expanded on a per subtree basis. Don't envy them but they brought it on themselves.
(, Fri 8 Feb 2013, 12:55, Share, Reply)
development has sadly ground to a halt,
the latest version with SMP has been in beta for frickin' years.
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 13:01,
Share,
Reply)
Really?
Ray tracing is pretty much the poster child for "SMP is easy". WTF have they been doing?
Actually I can guess. Globals all over the place instead of passing context around properly.
(, Fri 8 Feb 2013, 13:07, Share, Reply)
it seemed to work fine, when I tried it,
annoyingly they put a time limit on it though so you had to download a key to renew it every so often, for reasons I can't fathom. Fine if they actually expected it to come out of Beta within that time limit. Urgh.
I think it's because they added a lot of experimental features all in one go and haven't had the time to develop any of it.
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 13:13,
Share,
Reply)
Reminds me that I should get back into writing renderers.
I kind of got bored with it after I had been doing it for pay for a few years. I had a scheme to do a new hybrid MC/hardware renderer, but then got sidetracked and spent the last few years working on the tone mapping instead :)
(, Fri 8 Feb 2013, 13:20, Share, Reply)
hardware raytracing in FPGA pls!
I read a paper by a group who did that a few years ago, really impressive performance. As much parallel processing as you can shake a stick at.
It's funny how doing something for a living destroys your enthusiasm for it, though. I know this effect too well.
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 13:22,
Share,
Reply)
Problems with going the fpga route
1. fpgas are still pretty pricey
2. my current machine is a laptop, and /laptop/ fpgas are limited.
3. ray tracing needs random access to the whole scene. You either need to keep the whole scene in mem or have some paging system.
4. Data transfer is expensive. Whatever job you are offloading has to really be worth the effort.
3 and 4 are the biggies there. Interesting scenes are absurdly complex these days, and even gpu memory is limited. Not saying it can't be made useful, just that you need to think hard about what to offload.
Now if you offload the post render tone mapping you could probably do very well. I have been coding up a QMC based approx bilateral filter lately that gives excellent results and would work nicely in hardware.
When finished it's going to take pride of place in the middle of my detail enhancer. Same results as the old one, much faster :)
(, Fri 8 Feb 2013, 13:36, Share, Reply)
1. depends on spec. but they can be.
3 & 4 true, but, hmm... I wonder if a scene can be split up in some way and then composited afterwards.
(
Moon Girl Technologies horrendous beanbag, Fri 8 Feb 2013, 14:12,
Share,
Reply)
Yes, but with difficulty.
The problem is that after the first bounce a ray can end up anywhere in your scene, especially if you are doing lighting simulation or lots of reflections.
One approach is to split up the scene into regions and have a processor/machine/fpga dealing with each one, but then you have to deal with the problem of handing over rays between regions and routing results back to whatever asked for them.
The details get a bit horrible.
(, Fri 8 Feb 2013, 14:16, Share, Reply)
I am working on a new tool that uses light-touch RST to compress voxels into a manageable solid works T4T format.
(
feckler https://twitter.com/LeonardEctric1, Fri 8 Feb 2013, 12:54,
Share,
Reply)
Your user name is going to be confusing :D
(
Herb Alpert's Taxi Driver I have very beautiful lips, Fri 8 Feb 2013, 11:59,
Share,
Reply)
Only the other day, Fadge (no relation, honest), myself and Flirting With Badgers were all in the same thread. Badgergeddon.
(
FridgeBadger, Fri 8 Feb 2013, 12:16,
Share,
Reply)
someone has stolen your 16th c. Spanish lustreware and replaced it with a floral hat
(
Mighty Nibus who dares gins | @nibus, Fri 8 Feb 2013, 12:03,
Share,
Reply)
