Mtg 19/26: Tue-18-Mar-2025

Outline for Today

Sampling and Assignments

Administration

Today

  • a bit more about gaskets

For Next Meeting

Wiki

Link to the UR Courses wiki page for this meeting

Media

Transcript

Audio Transcript

  • Okay, so this is meeting 19. I
  • so I'm supposed to make A point of taking attendance today,
  • because the department head has had some reports of absenteeism.
  • So please do The attendance done? Okay? I'm
  • so I wanted to talk a little bit about the gasket from yesterday,
  • from last meeting, and I
  • so I it's about the bloat, as David called it. So what I was
  • doing is taking a sphere or a circle for and
  • it was centered here, oh,
  • moving the transforming it. I'm transforming this point halfway
  • here. Do that? Do
  • attendance. Yes, please, make sure you do the attendance. We
  • I just imagine those are all symmetrical. So the idea is, in
  • the limit, if we apply a bunch of transformations, we get a
  • little sphere. And
  • at the vertices of the triangle and at all the places in between
  • and the sphere is so we don't have to worry about the limit.
  • In the case of creating a desktop tetrahedron, we want to
  • make something that looks nice so you don't have to do Millions
  • of of spheres to get your Desktop ornament and
  • does that seem okay? I
  • so I've added some to the description for The desktop
  • tetrahedron assignment. So Start with The image and
  • so I've made a link to The Python code that I showed in
  • class. I
  • so this will produce a file that can be included in the PBR key
  • input file and
  • so we just have the include directive at the top of the
  • screen there and
  • so I recently discovered ew for my I
  • for water billing, and the leader is reporting. So now,
  • after I got a big bill, I've been I've logged in and set up
  • alerts and things, and now it's telling me that I'm using water
  • very substantially, and I'm trying to figure out what the
  • problem is. Your meters are
  • not screwy. That may be part of it. I
  • so you should take inspiration from the original Image done in
  • 1990 and feature a fractal tetrahedron at some level of
  • resolution.
  • So in using non diffuse materials, I
  • it, maybe gold or silver or glass,
  • and put it onto a desktop surface. Include more than one
  • light source, one of which is an area light source. You may add
  • other objects on the desktop surface to help demonstrate
  • shadows and other effects, including indirect lighting and
  • So when you see
  • that, Sometimes called color bleeding and
  • so here are Some examples
  • of the contribution of the of the
  • surfaces that aren't emitters themselves, but there's Some
  • energy coming off them so that color bleeds between the
  • surfaces and
  • The final image should be high resolution. So this, if it's if
  • it's square, it should be least 1024 by 1024 if it's not
  • squared, the smaller dimension should be at least 1024 I
  • five so I want you to explore the input options to generate a
  • high quality image. So consider the samples per pixel and the
  • sampler that you choose Use the stats option with pbrt and do
  • some analysis of the results with Image tool diff, for
  • example. So which combination of input parameter, input
  • directives, gives the Best image, and why so document the
  • pbrt file and
  • does That make sense? I
  • i Sorry it took me a while to find this.
  • So this may be a good place to look as Well, for Inspiration, I
  • is that C, VO, K, I,
  • so have six marks for creating the scene, recreating the
  • Desktop scene with the tetrahedron ornament on the
  • desktop, three marks for documentation and description of
  • what you've Done, and three marks for writing it with
  • different combinations of samples for pixels And samplers
  • and doing the analysis of which, if
  • we which set of input directives gives the Best image and
  • for the Advanced pbrt assignment, or I
  • so this is a little bit more detail, if It's helpful or not.
  • So pick a feature in pbrt version four, see, for example,
  • changes from version three, when they came so when they came up
  • in our class discussion, or you may find inspiration in the
  • exercises from the Book labeled with a two and explore its
  • impact. If
  • so maybe if you want to look at
  • some aspects of
  • the impact of focal length, different parameters for dealing
  • with the realistic camera and
  • exposure of the film
  • could do Something with Animated transformations. I
  • Does that seem okay?
  • Thanks, I'm just going to check something. I
  • so let's talk About our quiz from the
  • for today,
  • What's the best choice for post days, no purple poor folks. Yes.
  • And pre Egg is sinless.
  • So pre and post. So when we're talking about reconstructing an
  • image,
  • There are these two sources of quality problems and
  • so What is no discrepancy. Sadly, I
  • so with the example In the textbook,
  • something like This, I
  • I know, I think, we're often just thinking about the lower
  • corner. I
  • So based on this, if we have four points and one of them is
  • in the box, we can estimate the
  • size is one quarter.
  • But if this is point Three and point three and
  • Does that make sense? Does it match with your Understanding
  • from the textbook Example? Do
  • it. So codi sampling is the idea that we're trying to
  • take samples in a way that's reflective of your underlying
  • structure.
  • So third question, what is the last step in imagery
  • construction, put
  • the whole process in there for anyone that was paying
  • attention.
  • I i
  • gave you our courses sometimes I was I double checked that I had
  • deleted that I
  • so in this case, it's CS.
  • Yes, I
  • so what's the first step
  • Per PS, reconstruct, prefilter sample? Okay. I
  • So does this make sense?
  • So the image isn't can only be known by the samples that we
  • take of it and so
  • when we are
  • so that goes for when we're if we're trying to estimate The
  • value of pi with a Monte Carlo approach and
  • similarly, here we have so
  • the pixels as the samples of the function, and we Need to make
  • that into
  • Our image. So we're going to
  • I'm still bummed out about
  • putting the answer in the Question I
  • get over it. Does that make sense? I
  • so not focusing On the caption and
  • what do you think about The quality of this filter applying
  • a box filter. It's not very
  • good. So there are a lot of jagged inches here, not very
  • smooth. Okay, let's look at Gaussian filter.
  • So This is blurrier. I think I
  • still, really, jagged.
  • So if we look at this area here, and this area here,
  • it's a little smoother, I guess. Yeah,
  • I think it's still not
  • good, smoother, I Guess.
  • Okay, so let's look at third example and
  • would you say this one's better or worse?
  • Second one, I think it definitely works on the second
  • one. Colors are a little bit clear. I
  • plus weakness in this
  • one. So let's look at this area and
  • it seems to be a good mix of blurring and sharpness, So not
  • everything is blurred. Let's read
  • The Caption again.
  • So we can In looking at our samplers, we
  • you We can look at these filtering methods as well. I
  • so this is a box filter with a step function and with a set
  • side you saw Ital curve and
  • so it doesn't do a very Good job with a sign you saw, It'll
  • curve. I
  • so this is the Mitchell filter. You can see it dips into
  • negative a little bit here, so that helps to sharpen the image
  • And
  • so it does a Good job With Both I
  • that's white noise, and This is Blue noise. I
  • So
  • blue noise seems to lead to better
  • images and
  • So this is two different sampling patterns. One is
  • jittered sample,
  • jitter distribution, and the other one is Poisson desk
  • distribution. So what do you observe between the two of them?
  • Clustering? Yeah, there seems to be more clumping in A than B. B
  • seems to Be more evenly
  • distributed. I
  • So next we'll look at reflection models.
  • So we can develop reflection models by measuring data.
  • So measure data, and then we can also look at phenomenological
  • model, phenomenon,
  • biological models, we
  • equations that attempt to describe the qualitative
  • properties of real world surfaces can be remarkably
  • effective at mimicking them, and then We can do simulation and
  • and then we can use wave optics, so we have a detailed bottle of
  • light and
  • geometric optics. We've talked about a bit. I so
  • we
  • don't encounter diffraction with geometric optic models, so it's
  • A bit more straightforward and
  • so let's take a look at The first two sections of chapter
  • nine for Thursday. We
  • any questions or concerns I
  • Okay, thank you for Today, and we'll see You on Thursday. You
  • Thanks, you, too. You
  • do This work. I using fixation?
  • do This work. I using fixation?

Responses

No Responses