/3/ - 3DCG

No clue about them companies -- that's something I'd like to know about as well. Nevertheless, I do know a bit about the math.

The rendering equation is an integral equation and the material models are expressed as functions (or, more precisely, functionals), so some functional analysis would probably help if you want to deeply understand stuff. At least try looking into that direction, you don't have to understand much more than how it can be expanded and solved.

The rendering is usually performed using a stochastic integrator -- either plain old monte carlo or some metropolis-hastings method. Thus, it might make sense to read up a bit about stochastics, the most important things are probability measures and expectancy as integral on arbitrary spaces.

To see how these are applied to the rendering equation, perhaps the most important thing to read is the doctoral thesis of Eric Veach. It kind of set the grounds for these methods for the past 20 years.

You might also want to look into the original paper on the rendering equation by Kajiya and some of the later developments -- at least ERPT and primary sample space MLT are relatively simple yet powerful.

And implementing yourself is usually way more instructive than just reading or working on the abstract level.

>>466504
This guy. Surprisingly informed post for /3/

As for OP's question, I've never heard of actual mathematicians being employed by movie/animation studios or graphics R&D departments. They're still looking for computer science people, just with a solid background in mathematics and stochastics.

I would say the maths go much beyond basic trigonometry. Especially volumetric light transport digs deep into radiative transfer theory, and has close ties to research on neutron transport from nuclear physics.
Here's a small sampler of a recent paper that's a bit more on the math side, just to give you an idea:
http://www.ci.i.u-tokyo.ac.jp/~hachisuka/upbp.pdf

>>466504
>It kind of set the grounds for these methods for the past 20 years.

lol nope. He just collected other papers into one, and only for path tracing. Which isn't what we in the industry need right now. We need more animated content, faster - much faster - , with less actual employees - freelance or otherwise.

>>466523

Well, it was a good thread while it lasted, gents.

>>466529
>not giving raisins
>linking to nucleur physics and radiation when everybody knows that we need higher quality animations, faster, with less people much more so than a pixel looking 10% more accurate.

>>466529
Well at least he hasn't called anyone "kid". Yet.

>>466531
>"at least"
>still not giving any raisins

grow your ass up kid

>>466504
Said it better than I could have. Also OP, linear algebra is extremely important and useful for pretty much everything. And partial differential equations. If you want to do physics you need that shit, and if you want to do rendering you want to do physics.

>>466533
>And partial differential equations. If you want to do physics you need that shit, and if you want to do rendering you want to do physics.

...none of this is important in the least

>>466535
Unless you want to do some cool volume preserving animation for example.

>>466537
Just paint in blend shapes for pose space fixes

>>466538
That's the traditional solution, but there are better methods. Guess not in realtime though.

>>466551
>there are better methods

not really. You make up some big as hell muscle system but you can just paint your pose space blends in a few minutes like they have in the new Modo or for years in maya with comet / joe alter PSD.

I see. Is rendering the most popular field of research? How about simulations?

>>466745
It's not at all. Physically based simulation is even more dead. Both fields have been shrinking for the last few years, with a lot of people switching to more "hip" areas like fabrication or computer vision.

>>466756
>Physically based simulation
>Physically based...simulation

If a simulation isn't based on our physical world, what the fuck would it be based on you fucking underage troll

>>466762
by simulation he means something that actually tries to be correct

most of the time you just fib it to get something that's roughly accurate, but the method may not be similar to what actually happens

>>466832
>by simulation he means something that actually tries to be correct

>simulate
>something....incorrectly....

>>466762
Jesus Christ man learn some Computer Graphics terminology. "physically based" is not exactly an obscure term.

Other fields like engineering are doing "physically correct" simulation, or, well, just "simulation". Their goal is to arrive at the correct result, and incorrect results are useless to them.

In computer graphics, we put emphasis on "physically based". The methods developed here are based in physical principles (usually the same PDEs as other fields), but are simplified and approximated. The main goal is speed while still looking visually plausible. The goal is not to compute the correct result, since artists don't care about that. Most simulations in CG you will encounter are not physically correct at all - there are no bounds on the error, no convergence measures, it just looks plausible to the human eye. Hence "physically based", not "physically correct".

On the opposite end of the spectrum, there's all the ad hoc methods made up by TDs or artists that are not physically based at all. These methods do not start with physical principles and simplify until they're fast enough, they're just completely made up from the start and tweaked until they look good. This is what differentiates e.g. Toy Story era shaders from physically based rendering.

Go pick up a book.

>>466762
>>466834
You're the retard bro, you can simulate any range of variables on a computer without being anywhere close to physically correct.

Quick example; you could simulate how it is to be on a world with say negative friction, where brushing up against a surface causes your velocity to grow.
There is no such thing in nature so it's physically incorrect, but it is still a simulation. Or in a game where you give incorrect values for many variables
intentionally to give the physics a certain feel.

So take all of your reaction images anon and shove them so far up your ass that they hit the top kek of your thought wreck.

>simulations are dead
Then what are all these papers about?

http://mmacklin.com/pbf_sig_preprint.pdf

http://mmacklin.com/uppfrta_preprint.pdf

http://matthias-mueller-fischer.ch/publications/strainBasedDynamics.pdf

>>466552
Yes really, nobody doesn't bother implementing newer methods because fixing everything by hand is already the standard and works. But I guess it'd be cool to save on the manual labor.

>>467119
The term "dead" was probably used to highlight the contrast when it used to be a major part of CG research. A bit like rendering, which is not at all as lively as it used to be.

>>466498
>What kind of work do they do?
Any work you want
30k starting

>>467119
This is beautiful. Has anyone here developed anything with PhysX SDK?

>>467613
yick, closed source nvidia bullshit. Wont even run on amd.

>>467614
>closed source nvidia bullshit
https://developer.nvidia.com/physx-source-github

>Wont even run on amd
SDK runs on any setup, except for Nvidia CUDA stuff.

>>467623
>https://developer.nvidia.com/physx-source-github

thats worthless cpu only, kid. Physx does not run on AMD.

The only thing to use is Bullet 3

>>467624
How is CPU only worthless? Keep in mind we're talking about simulations, so vidya related reasons don't apply here.

Also, AMD needs to fucking step up.

>>467625
It's beyond worthless, especially with modern, large sims and increasing gpu card power. When pascal hits next year and we get 10x speed increase like huang said two days ago at the conference, we'll finally get better realtime sims, all on gpu. This cpu shit, its laughable.

Check Euclideon hyper-realistic engine using points instead of polygons.

http://www.youtube.com/watch?v=DbMpqqCCrFQ
http://www.youtube.com/watch?v=00gAbgBu8R4

>>467664
Oh god, looks like we have another gullible kid here. I hate to break it to you, but that shit is useless snake oil. He only made those to get funding from the local government to support his shitty business.

Point clouds are an incredibly inefficient way to simulate a world. It requires terabytes of storage space compared to polygons and texture maps, it cannot be deformed nor animated anywhere near as efficiently, and it is much harder to control the surface look artistically.

Those videos look good because they are no different from fucking photographs. The lighting is static for the most part, just like a photo, it's a 3D scan of an actual location, with all that baked in real-world lighting to make it actually look good.

>fucking photographs
Yeah, nobody likes photographs, especially walking around inside one.

This thread died rather fast. Unfortunate too, I was hoping to learn something from it.

>>467721
It's not new technology for that purposes you dunce. Interactive point clouds are nothing new and is what his business and many others already do. It's nothing ground breaking and it certainly won't be used for games as he claims it's perfect for. That's the snake oil I'm talking about, people think because it looks so realistic compared to games but don't realize you can already do the same shit with polygons if you want as well and walk around in the 3D scanned world, the problem is it's just not fucking useful for anything other than visualization of existing land.

>>468084
Thanks for the links to equally photo-realistic polygonal renders.

>>466504
How would you suggest implementing it yourself? Building a renderer in OpenGL? C++? The math in the paper seems manageable, but I'm pretty ignorant otherwise.

>>468324
Just wait for Vulkan, jump in and write a good article asap.

Pixar has some good papers to give you an idea of what a mathematician could delve into when it comes to CG.

http://graphics.pixar.com/library/indexTechmemos.html

I guess if you had something published in that area it wouldn't be too far from getting a job in a company like this. I'm assuming you can program well (my experience with mathematicians is that generally they can code pretty well) so that shouldn't be too big of a problem as long as you're not a physicist.