Search found 83 matches

by MohamedSakr
Sat Aug 15, 2015 8:01 am
Forum: General Development
Topic: gradient domain path tracing (GPT)
Replies: 10
Views: 12817

Re: gradient domain path tracing (GPT)

ok, after some googling, tracing code line by line, I can understand the code to some extent and convert the code "as it is" to C++ now the function takes mainly a single matrix X, function [Z, obj]= LRR_IRLS(X,lambda,rc,rho,display) where I can imagine X here can represent v_p. so how to use the gr...
by MohamedSakr
Sat Aug 15, 2015 2:47 am
Forum: General Development
Topic: gradient domain path tracing (GPT)
Replies: 10
Views: 12817

Re: gradient domain path tracing (GPT)

Hi Bachi, thanks for clarifications, this really helped :) I understood what you said. as a continuation, hope you may clarify if this paper "may" help in solving the problem efficiently. the author page: https://sites.google.com/site/canyilu/ paper: Smoothed Low Rank and Sparse Matrix Recovery by I...
by MohamedSakr
Fri Aug 14, 2015 3:54 pm
Forum: General Development
Topic: gradient domain path tracing (GPT)
Replies: 10
Views: 12817

gradient domain path tracing (GPT)

2 days ago I found this online, https://gist.github.com/BachiLi/4f5c6e5a4fef5773dab1 tested it, works pretty well, but my mathematical experience couldn't allow me to modify the resultant poisson reconstruction much. from the results, some how they are matching L2 minimization, how to do L1 minimiza...
by MohamedSakr
Tue Jul 21, 2015 4:01 am
Forum: General Development
Topic: Multiplication between two slerped quaternion
Replies: 7
Views: 5829

Re: Multiplication between two slerped quaternion

I see :D
thanks for clarification

considering "correct" interpolation, I'm still aware of why you think that multiple slerp is correct?
its transform(t) is proportional to t^n which seems wrong to me
by MohamedSakr
Mon Jul 20, 2015 8:30 pm
Forum: General Development
Topic: Multiplication between two slerped quaternion
Replies: 7
Views: 5829

Re: Multiplication between two slerped quaternion

I meant linear interpolation, so why it is wrong?
by MohamedSakr
Mon Jul 20, 2015 8:25 pm
Forum: General Development
Topic: pbrt-v3 source code now available
Replies: 23
Views: 24798

Re: pbrt-v3 source code now available

I think at these lines: https://github.com/mmp/pbrt-v3/blob/6b6c30f1e2bdb0052aed0136638d7c6106de779d/src/core/film.h#L130-L152 for loop should be using <= so if we are at pixel_x = 11, filter radius = 0.5 xmin = 11 - 0.5 - 0.5 = 10 xmax = 11 - 0.5 + 0.5 + 1 = 12 loop from 10 to 12 (int i = xmin; i <...
by MohamedSakr
Sun Jul 19, 2015 1:30 pm
Forum: General Development
Topic: Multiplication between two slerped quaternion
Replies: 7
Views: 5829

Re: Multiplication between two slerped quaternion

It seems that Slerp(M1, M2, t) * Slerp(M3, M4, t) is not equal to Slerp(M1M3, M2M4, t). let's consider both cases and check which one is true :D : a) Slerp(M1, M2, t) * Slerp(M3, M4, t) M = (t * M1 + (1 - t) * M2) * (t * M3 + (1 - t) * M4); //this looks wrong, as M will be proportional to t^N, wher...
by MohamedSakr
Sun Jul 19, 2015 12:25 am
Forum: General Development
Topic: Multiplication between two slerped quaternion
Replies: 7
Views: 5829

Re: Multiplication between two slerped quaternion

I believe that when you want to render a complex scene, everything should be "flat"
so convert your hierarchy scene as a pre-process to a flat scene "this would take a fraction of a second"
by MohamedSakr
Wed Jun 17, 2015 5:33 am
Forum: General Development
Topic: Float versus double
Replies: 12
Views: 14783

Re: Float versus double

I would prefer floats, at least when you want to port your code to GPU it will be much easier, also floats are cheaper in memory operations, and better for SIMD. precision issue isn't a real problem, most of the time in "real scenes" , there are options to generate bias "ray epsilon, sample intensit...