Search found 46 matches

by apaffy
Tue Aug 21, 2012 9:40 pm
Forum: General Development
Topic: MIS BPT with Russian Roulette
Replies: 16
Views: 13369

Re: MIS BPT with Russian Roulette

I tried to check uniform weighting (in this case, there are 3 strategies [s=0, t=4], [s=1, t=3], [s=2, t=2], they are assigned 0.33333... weight). And I got the result, which shows that BPT and reference will converge to the same. From this result, I think that my MIS weights computation part is in...
by apaffy
Tue Aug 21, 2012 1:50 pm
Forum: General Development
Topic: MIS BPT with Russian Roulette
Replies: 16
Views: 13369

Re: MIS BPT with Russian Roulette

I simply tried to comment out multiplying and dividing by RR in MIS computations. (not in subpath traceing) But the result yet seems incorrect:? LuxRender includes BPT implementation and it seems to consider RR in MIS weight computations. Then I think you have to conclude that the remaining MIS cod...
by apaffy
Tue Aug 21, 2012 12:31 pm
Forum: General Development
Topic: MIS BPT with Russian Roulette
Replies: 16
Views: 13369

Re: MIS BPT with Russian Roulette

It's difficult to find a bug in the code just by looking, but I thought I'd comment on this: Of course, I know that it is required that the probability for generating a adjacent vertex in MIS weight computations contain RR continuation probability. Actually, as Dietger noted on another thread, you d...
by apaffy
Tue Aug 21, 2012 12:24 pm
Forum: General Development
Topic: MIS on the GPU
Replies: 41
Views: 27266

Re: MIS on the GPU

The solution of splitting the BSDF into reflective and transmissive parts does mean some paths will be zeroed when using shading normals. In extreme cases these can produce black patches. I'm not aware of a perfect solution to this problem. Ignoring the issue completely will result in light "leaking...
by apaffy
Fri Aug 17, 2012 5:20 pm
Forum: General Development
Topic: MIS on the GPU
Replies: 41
Views: 27266

Re: MIS on the GPU

that s right ! but then what is the cos term in figure 5.8 , it is not in my formula ? Figure 5.8 is an implementation of a kernel defined by eqns 5.20 and 5.21, which merges together the BSDF and the pdf conversion (from solid angle to projected solid angle). This is what I was talking about earli...
by apaffy
Fri Aug 17, 2012 3:51 pm
Forum: General Development
Topic: MIS on the GPU
Replies: 41
Views: 27266

Re: MIS on the GPU

A new question please, in Figure 5.8, when applying the Adjoint BSDF it returns f * |Wo . Ns|, where Wo is toward the light... So, my equation here is incorrect ? I should not use cos(theta_0)=|(x0->x1).N0| and cos(theta_1) =|(x1->x2).N1| but |(x1->x0).N0| and |(x2->x1).N1| instead ? I think you're...
by apaffy
Fri Aug 17, 2012 12:32 pm
Forum: General Development
Topic: MIS on the GPU
Replies: 41
Views: 27266

Re: MIS on the GPU

It seems more coherent compared to my path tracer... but... I'm surprised that the BSDF is expressed in term of path-tracing and not in light-tracing. Do you know why it is not expressed in the physical direction of light ? Which seems more natural ! Something is confusing , in Figure 5.8, when adj...
by apaffy
Fri Aug 17, 2012 9:08 am
Forum: General Development
Topic: MIS on the GPU
Replies: 41
Views: 27266

Re: MIS on the GPU

Shading normals are tricky to get right! 1. The adjoint BSDF is for light subpath construction (so either sampling importance or simulating light particles). The BSDF is for eye subpath construction (so either sampling radiance or simulating adjoint photons). 2. I find it easier to compute the scali...
by apaffy
Mon Aug 13, 2012 6:49 am
Forum: Visuals, Tools, Demos & Sources
Topic: Topmod!
Replies: 3
Views: 6327

Re: Topmod!

Zelcious wrote:Solid or hollow?
Solid glass, using a fresnel dielectric BSDF. IOR is around 1.5, the glass medium has (RGB) absorption but no scattering.
by apaffy
Sun Aug 12, 2012 9:33 pm
Forum: Visuals, Tools, Demos & Sources
Topic: Topmod!
Replies: 3
Views: 6327

Topmod!

Topmod seems to be unchanged from 2007 but it's still fun to play around with: http://code.google.com/p/topmod/ topmod_glass.jpg Image rendered using bidirectional path tracing. I'm not sure the image is completely correct, the high curvature bits look a bit dark to me and could well indicate I have...