He is talking about geometric optics, bidirectional reflective distribution functions, chemical structure of materials, optimisation for both probability models and for computer hardware. Thats mathematics.
Very interesting. Despite going to college for a different degree (CS), it seems that the problem solving needed is very similar. Very enlightening video!
Those are great tutorials, true. But arcsynthesis is not outdated and goes into details behind the mathematics which really helps why you try and troubleshoot problems that you will invariably run into. TLDR use both :)
I notice he didn't mention the hemicubes method for radiosity which largely solves the occlusion and visibility problem, but does it via rasterization using the same methods rasterizers have at their disposal for dealing with visibility. But in all fairness, it seemed like he was using that to segueway into MCRT. I've done talks like this one and if you start including every solution people have used, it's very hard to fit it into a definite time slot.
He mentions that raytracing is slower still, which is semi-inaccurate. Currently, what I'm seeing in my work in offline CG is that raytracing generally beats rasterization, partly because of trivial parallelization, but also because you don't need all the extra passes and data I/O for shadows/GI. It's slower if you just look at shade time only, but when you add up all the extra work a conventional micropoly rasterizer has to do to support all those features, it works out faster in practice.
Google 'arcsynthesis tutorials' for guides on getting into opengl. Or google 'udacity rendering' for a free video course on this kind of stuff (from first principles).
programming OpenGL in C++ let you use the graphics card, then it's up to you to bash your head for years to gain a fine result as they do in their current tech :D try to learn with ogldev.net, good and easy tutorials if you know c++
I felt great already knowing everything he was talking about, made me feel for a moment as brilliant as the guy, just for a moment.
He is talking about geometric optics, bidirectional reflective distribution functions, chemical structure of materials, optimisation for both probability models and for computer hardware.
Thats mathematics.
Very interesting. Despite going to college for a different degree (CS), it seems that the problem solving needed is very similar. Very enlightening video!
Those are great tutorials, true. But arcsynthesis is not outdated and goes into details behind the mathematics which really helps why you try and troubleshoot problems that you will invariably run into. TLDR use both :)
I notice he didn't mention the hemicubes method for radiosity which largely solves the occlusion and visibility problem, but does it via rasterization using the same methods rasterizers have at their disposal for dealing with visibility. But in all fairness, it seemed like he was using that to segueway into MCRT. I've done talks like this one and if you start including every solution people have used, it's very hard to fit it into a definite time slot.
He mentions that raytracing is slower still, which is semi-inaccurate. Currently, what I'm seeing in my work in offline CG is that raytracing generally beats rasterization, partly because of trivial parallelization, but also because you don't need all the extra passes and data I/O for shadows/GI. It's slower if you just look at shade time only, but when you add up all the extra work a conventional micropoly rasterizer has to do to support all those features, it works out faster in practice.
k... so how do you implement this into the computer and make the magic work.....
Google 'arcsynthesis tutorials' for guides on getting into opengl. Or google 'udacity rendering' for a free video course on this kind of stuff (from first principles).
Arcsynthesis is outdating. Just go to open.gl.
whole 0 people disliked this video because it was to difficult to understand
I hate SSAO
programming OpenGL in C++ let you use the graphics card, then it's up to you to bash your head for years to gain a fine result as they do in their current tech :D
try to learn with ogldev.net, good and easy tutorials if you know c++