Mtg 4/26: Thu-16-Jan-2025

Outline for Today

Monte Carlo

Administration

Media

Transcript

Audio Transcript

Okay, meeting four of 26 last meeting, we focused on
building pbrt version Four
and some wiki stuff. Any questions about that? Applause.
So I realized I set up a scheduler too late and it won't
let me. It didn't let me set up a time for a slot for the first
meeting. So Laura, you can remove your if, David, you're
going to do yesterday's or last meeting?
For last meeting, yeah, okay,
because more than one person can do,
connect it, right? Yes. Everyone can edit I'm just and nobody
else has signed up yet,
so I'm going to send you a reminder to sign up. So thoughts
about the wiki pages. So on the wiki pages add navigation. So I
did that on first meeting, and then once you have a page
edited, posted to the class discussion forum, and then I
will make The link to the meeting page,
which right now doesn't do anything.
Right now does this because I don't have the ID, the page Id
set up. So if you post it, then people know it's there, and then
I'll get the information and make the link and
and I fixed the attendance.
So attendance, the link in the meeting goes to the proper you
are Courses page and
so I left the link for The wiki documentation up. So our ta
xylin, did we? Windows build of PBR T version four, and he
posted his comments here.
In case you don't recognize that link, that's the one that I
thought was not very good, but it appears it's better than,
better than I first gave it credit for. Most of the
procedures in the video are correct. So that's the first
thing to look at. You may need to change the command C make.
So the difference would be that minus a x 64 is removed from the
second one, and then there's a Pull Request that talks about
dealing with the P techs error. So if that comes up for you,
that might be a helpful
thing to do. So any progress on
mine was the version I was trying to compile using the
wrong version. So I need a version 16 for the Visual Studio
I have on my computer. Version 17. I
went back. Tried that. It worked, just compiling it. Okay.
So did you get I'm close. Did you get Visual Studio from? I
had Visual Studio already on my computer. I didn't know what
version I had, so I was able to from that video that you showed,
I saw where he found the version, and I went in mind and
realized that my version is version 16, Visual Studio. So
that was the year that I was getting when I was trying to put
it all together. It was trying to use Version 17, and it didn't
exist on my computer, which is true,
version 17 did not exist. So can you update? I'm
able to use. I was able to get it to work with, okay,
so Visual Studio at all?
Okay, so I was able to get it because I just had to change the
version in the compilation or in the setup and did everything. It
was closed, okay,
and see make, was that working for you? Yeah,
I got CMake working. I have a compiled thing of pbrt Four. I
not test if it works. I realize that should have been an
important step. But if I can open it, and it opens, I got
things to work, probably
closer. Let's go with that,
pretty sure compiled.
If you can open it, I would say, yes,
I just haven't rendered anything, is all. Yeah, that's
it took forever for me to download the scenes. So once I
got that first part working, I downloaded the scenes. I just
haven't done a rendering of the city of the scene.
Okay? So was there any when you got C make to work? Were you
able? Did you notice, anything you needed to change?
I was having trouble with the Z lip, static step, and my buddy,
he was getting I had a friend who walked on to help me with
it, because he was, he's not in this class at all. He just knows
this kind of stuff. But so he was trying to help me, and he
ended up bypassing that step by giving me a already installed
copy. Somehow, I have no idea what he did, but He sent me a
working copy. Okay?
So yes,
Bless you. I
Okay, so
any other comments about this before we move on? I It would be
nice if you have had success, share it with the class and The
class Discussion Forum. I
I'm not happy the way I started that sentence. I
I'm paraphrasing the so we have a lot of integrals to deal with,
and we can't do An analytical solution. So how do we handle
them. I
Anything else to add a
pretty decent approximation of integral through just random
Walk analysis, like Monte Carlos, Just over, Over, Over,
CS,
I supposed To do rendering I
probability mass function.
Yeah, I I just I was curious. Instead of downloading, going
back to pbrt for a second, instead of downloading the
repository with the with the PBR TV for scenes, is there an
example? Because that video that I disparaged it, talked about
having a PBR key file in the project.
It might be somewhere in there. It's just, it's a very simple
scene. So maybe it's even a better indicator on whether or
not you can run it. Is if you can run the simple scene,
because it was just like a circle right over, yeah,
sphere and a checkerboard, yeah.
So that might be a sufficient indicator whether You can run
it. If it's in there you
so It might be available in the project repository, you can copy
and paste it from pbrt.org file
format
we looked at that the other day. Okay, yes, that
bottom line says, suited to the integrals, that something up
with rendering had
Come up With rendering. I
so they use Las Vegas algorithms. Use random, some
randomization, but they always come up with The same answer You
okay, so Can
Are you more familiar with Las Vegas or Monte Carlo? Monte
Carlo as the destination for gamblers.
So I was reading about this, and there's one, one source I found
that said Monte Carlo was named after the name was chosen
because of Someone's uncle like to go To the casinos there. I
I would take a numerical analysis.
So I don't want to make this about numerical analysis. I want
to give a bit of context so we can understand how to use the
software. Okay, so fancy integrals will not be part of
the exams. I
You, maybe for you, we can no
ignore him. He's speaking facetious.
Direction.
So what? What is the idea behind integration? Generally,
if you have, like a curve to figure the area under a curve,
yeah, you don't know an exact, like, straight angle, except
for, like a derivative, you add up a whole bunch of little
shapes underneath that curve, until closer and closer and
closer it gets To being that curve essentially With the
estimated Pi over
so quadrature comes up, and that's based on fitting squares
under the curve.
Instead, it wasn't very valid for this because it never really
converged on the higher dimensional kind of things, kind
of, yeah,
so these exactly
saying, like, three dimensionally, you can't use to
collect quantum sure, because you'd be using cubiture At that
point, right? If you're doing three dimensional scenes. Is
that what it means?
So I think these are good for, smooth functions, and functions
we're dealing with aren't always Smooth and they're not disjoint.
So the advantage of The
I'm not sure how to write this with the error negligible.
Negligible.
The areas like so minute it's important.
Yeah, it increases
exponentially as sample size increases exactly.
I don't want to hook up. Writing is in the wiki To the pages I've
linked you.
It reminds me story about when I went to a Star Trek convention,
they said we talked about how script they have Star Trek, but
they said the script writer did something, and then they put
blanks like that for the science consultant to fill them in.
Our choice is negligible. You
so Here's the idea of Monte Carlo integration. So in this
case, we have a circle inscribed with a square, and then we do
we plot, we randomly generate points, and we calculate how
many are inside the circle and How many are outside we
so that's a way to get an estimate of pi and
so here are a few versions of code to do the I
estimation of pi.
So there's C Plus Plus.
Mathematica You
and Here's an animation of The pie. Example i
Does that make sense? So
so the question is, how do we make The most of the samples
that were given the
so we can't always add more to compute an image. I
So what are some approaches
take more efficient samples. So if you have a phenomena where,
if you know accuracy, increases by a distribution of where you
put those samples, you presume that it follows a distribution
to begin with as you sample it, so you don't need as Many
samples to get the same level of variance. Do
so I'll show an example from
the book Question.
I So here's the Stanford bunny.
This was
an illustration in chapter two, and then there's also i
So this was the illustration for more efficient sampling, because
The computation these were done with the same computation, same
amount of computation, but the one On the right here is using a
different sampling the
and it's not easy to see With that,
but there's a tool in the repository and I'll
Put
these images on into the media for the
for this meeting, but here I am running
this on my laptop, so I took the Save the screen took screenshots
of the images, and that's why they have these ridiculous
04am and screenshot from
Tuesday at 1103 and so
this image tool diff, so I'm Using the first image as a
reference, and then comparing it with the second one.
So we've gotten a change
of point zero 2% and
So that's helpful. I
you can see the edges are i
It's where the difference occurs most pronounced. But you can see
there's differences On the surface as well. I
so the idea is, you want To do
Focus soundly and
does that make sense?
So would that be like, saying, If you sample up here, and then
you sample down here, and that's like a major difference, you do
a lot of sampling In this area, because there was the major
offset. I
I wonder, I guess The short answer is, Yes. I
it so we might look at the variance.
Are there more minions? Do you think it's gonna more minion
movies coming along? It seems like they've gotten all they can
out of that. I
I think they've gotten all they can now The Terminator movies
too, But they seem To I
so if We've Got A
That's The
so I made the image bit bigger, like
five times i
Here it is. I Send It
so I shouldn't have changed the size of the image. So
Let's do It def.
Okay. Thank You.
So that didn't change much. I
let's try one More revision
here. I I thought someone was trying to get into
this. That's A trigger difference. I
do You still want more noise?
Yeah, so I reduced the samples to 16,
and that also created
the surplus of pixels that are out of Ganon. So when I did 128
samples at the resolution of 768 I
had one Gan pixel. And then when I did the 16 samples, There are
33,490 out of gallon pixel channels. I
so we're at a time I will
do some Work on getting exploring the parameter spaces
for this simple PBR key file. And we'll start on chapter
three. Well, we'll give you the questions that come up. Get into
Chapter Three next week. Applause, thanks for your time
today. Take care of a good weekend. Stay warm.

Responses

No Responses

Site built: 2025-02-21 17:51 -0600 • Give feedback about page

Mtg 4/26: Thu-16-Jan-2025

Outline for Today

Monte Carlo

Administration

Today

For Next Meeting

Wiki

Media

Transcript

Audio Transcript

Responses