Timeline for Multiple Importance Sampling in Path tracer produces Dark Images
Current License: CC BY-SA 4.0
33 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 24, 2019 at 19:47 | history | edited | gallickgunner | CC BY-SA 4.0 | added 131 characters in body |
| Feb 24, 2019 at 19:46 | vote | accept | gallickgunner | ||
| Feb 24, 2019 at 19:41 | history | edited | gallickgunner | CC BY-SA 4.0 | added 278 characters in body |
| Feb 24, 2019 at 17:59 | answer | added | lightxbulb | timeline score: 4 | |
| Feb 24, 2019 at 15:49 | comment | added | gallickgunner | @lightxbulb - Check now. | |
| Feb 24, 2019 at 15:49 | history | edited | gallickgunner | CC BY-SA 4.0 | added 1328 characters in body |
| Feb 24, 2019 at 15:25 | comment | added | lightxbulb | Note that I have gone through your code, and I believe there are a number of mistakes. I think that if you do a small overview of the relevant parts then: 1) I'll be certain it's not me misunderstanding what your code does, 2) I'll be able to explain why theoretically what you are doing is incorrect (without talking about the implementation specifically). I just think you've misunderstood some parts of the derivation of MIS. If we can clarify which parts, then I can do a detailed writeup. | |
| Feb 24, 2019 at 15:08 | comment | added | lightxbulb | @gallickgunner I saw that you linked the code, however, I'd rather see the theoretical explanation of what you're doing rather than go through the code. Just a high level overview is ok. | |
| Feb 24, 2019 at 14:45 | comment | added | gallickgunner | @lightxbulb - I actually posted the link to the whole code incase you overlooked it. Check from line 471. Didn't want to add here since it's gotten messy and long as it is. The weights are computed using power heuristic (1 line function at the end of the file). If you still want I can try posting here as well. | |
| Feb 24, 2019 at 14:17 | comment | added | lightxbulb | Could you add a small LaTeX derivation as to how you compute your weights, and at what events you do so, and the the final estimator you use. Basically a small explanation as to when and how you do direct light sampling, and brdf sampling and then the way you combine those. I can also link you some code of mine with MIS in glsl if you want. | |
| Feb 22, 2019 at 13:13 | history | edited | gallickgunner | CC BY-SA 4.0 | added 196 characters in body |
| Feb 20, 2019 at 14:07 | history | edited | gallickgunner | CC BY-SA 4.0 | added 17 characters in body |
| Feb 20, 2019 at 11:21 | history | edited | gallickgunner | CC BY-SA 4.0 | added 168 characters in body |
| Feb 20, 2019 at 9:46 | comment | added | gallickgunner | I need to update this. Should I edit out the previous stuff or leave it as it is so people can follow the changes? | |
| Feb 20, 2019 at 9:21 | history | edited | gallickgunner | CC BY-SA 4.0 | deleted 9 characters in body |
| Feb 20, 2019 at 1:00 | history | bumped | CommunityBot | This question has answers that may be good or bad; the system has marked it active so that they can be reviewed. | |
| Jan 21, 2019 at 0:32 | history | edited | gallickgunner | CC BY-SA 4.0 | added 9 characters in body |
| Jan 19, 2019 at 22:00 | history | edited | gallickgunner | CC BY-SA 4.0 | deleted 9 characters in body |
| Jan 15, 2019 at 9:34 | comment | added | gallickgunner | @trichoplax - If i removed it, this gets a little messy. This is how I do it. I'm now accumulating color from two light sources but it's a single sample ( we don't want to divide by 2). However for every light source I also take a brdf_sample. This accounts as 2 BRDF samples. Hence I use MIS with the formula accounting for 1 light sample and 2 BRDF samples. Result seem almost same to me. Note if i assume those 2 brdf samples as 1, results are very bright, like way bright so I think the first approach is right. Updated github so you can see how I did it. | |
| Jan 14, 2019 at 22:38 | comment | added | trichoplax is on Codidact now | What do you get if you take out your light source choosing heuristic? | |
| Jan 14, 2019 at 9:46 | history | edited | gallickgunner | CC BY-SA 4.0 | added 364 characters in body |
| Jan 14, 2019 at 9:08 | history | edited | gallickgunner | CC BY-SA 4.0 | deleted 404 characters in body |
| Jan 14, 2019 at 2:07 | comment | added | PaulHK | A shot in the dark really, I have seen other question on here were the solution was surprisingly that. | |
| Jan 11, 2019 at 10:33 | comment | added | gallickgunner | @PaulHK - What would that imply? I don't think it's actually PI times darker, its just mis_weight times darker/brighter. Because as I said as soon as i remove mis_wieght and just divide by light_pdf I get the brighter image. | |
| Jan 11, 2019 at 7:38 | comment | added | PaulHK | May be a coincidence but are your diffuse surfaces exactly PI times darker ? | |
| Jan 10, 2019 at 10:18 | history | edited | gallickgunner | CC BY-SA 4.0 | deleted 1 character in body |
| Jan 8, 2019 at 14:46 | answer | added | sriravic | timeline score: 2 | |
| Jan 7, 2019 at 9:57 | comment | added | gallickgunner | Im using cosine weighted hemispherical sampling so the BRDF PDF is supposed to be cos(theta)/pi | |
| Jan 7, 2019 at 6:26 | comment | added | Hubble | You shouldn't include the cosine foreshortening term in your brdf_pdf. | |
| Jan 5, 2019 at 16:38 | history | edited | gallickgunner | CC BY-SA 4.0 | added 348 characters in body |
| Jan 5, 2019 at 16:25 | history | edited | gallickgunner | CC BY-SA 4.0 | added 69 characters in body |
| Jan 4, 2019 at 10:54 | history | edited | gallickgunner | CC BY-SA 4.0 | added 59 characters in body |
| Jan 4, 2019 at 2:20 | history | asked | gallickgunner | CC BY-SA 4.0 |