I asked about during the Q&A session but by that point many people had left, but I thought the answer still interesting enough that documenting it is good. Keenan mentioned that this is actually a used technique in terms of rendering fluids in practice: that is, do a physical simulation on some finite set of particles, then do this sort of surface blending to create a sort of volumetric model of the fluid. He sent this paper which has a good example of this technique in action.
idontknow
Thanks for posting this here. That's definitely an interesting example for implicit surface representation. When I was watching this lecture I had that question too
0x484884
I'm a little confused with what's going on in the 2d example. It looks like the plane is fixed and then we're moving the centers of the gaussians towards each other but I don't understand relation of the pictures and the values of f.
0x484884
Also it seems natural to have the total area in 2d (or volume in 3d) to remain constant as the shapes merge but this doesn't seem to be the case here. Is there a way to change the method, such as changing the value the gaussians must sum to over time to make this the case?
@0x484884 Yeah, that's a minor bug in the slide. The pictures are actually showing the points move around, rather than changing the height of the plane (which will have a similar effect).
I asked about during the Q&A session but by that point many people had left, but I thought the answer still interesting enough that documenting it is good. Keenan mentioned that this is actually a used technique in terms of rendering fluids in practice: that is, do a physical simulation on some finite set of particles, then do this sort of surface blending to create a sort of volumetric model of the fluid. He sent this paper which has a good example of this technique in action.
Thanks for posting this here. That's definitely an interesting example for implicit surface representation. When I was watching this lecture I had that question too
I'm a little confused with what's going on in the 2d example. It looks like the plane is fixed and then we're moving the centers of the gaussians towards each other but I don't understand relation of the pictures and the values of f.
Also it seems natural to have the total area in 2d (or volume in 3d) to remain constant as the shapes merge but this doesn't seem to be the case here. Is there a way to change the method, such as changing the value the gaussians must sum to over time to make this the case?
I made the 2d version in Desmos to help me visualize this better. Its pretty fun to play around with. Heres the link: https://www.desmos.com/calculator/lnaay9c1g4
@0x484884 Yeah, that's a minor bug in the slide. The pictures are actually showing the points move around, rather than changing the height of the plane (which will have a similar effect).
@0x484884 Nice little Desmos demo, btw!