Dr. Zsolnai covers a new computer graphics method that simulates soap bubbles with realistic detail: interference rainbow patterns, how they deform and rupture, and swirling surface behavior. The simulation adds 4 to 31 percent computational cost depending on complexity, and matches real footage closely.
Dear fellow scholars, this is Two Minute Papers with Dr. Zsolnai. Today I will try to show you the incredible progress in computer graphics research through the lens of bubbles in computer simulations. Yes, bubbles indeed. Approximately a year ago, we covered a technique which could be used to add bubbles to an already existing fluid simulation. This paper appeared in 2012 and described a super simple method that helped us computer where bubbles appear and disappear over time. The best part of this was that this could be added after the simulation has been finalized, which is an insane value proposition. If we find ourselves yearning for some bubbles, we just add them afterwards and if we don't like the results, we can take them out with one click. Now, simulations are not only about sights. What about sounds? In 2016, this paper did something that previously seemed impossible. It took this kind of simulation data and made sure that now we can not only add bubbles to a plane water simulation, but also simulate how they would sound. On the geometry side, a follow up paper appeared just a year later that could simulate a handful of bubbles colliding and sticking together. Then three years later, in 2020, Christopher Betty's group also proposed a method that was capable of simulating merging and coalescing behavior on larger scale simulations. So what about today's paper? Are we going even larger with hundreds of thousands or maybe even millions of bubbles? No. We are going to take just one bubble, or at most a handful, and have a real close look at a method that is capable of simulating these beautiful evolving rainbow patterns. The key to this work is that it is modeling how the thickness of the surfaces changes over time that makes all the difference. Let's look under the hood and observe how much of an effect the evolving layer thickness has on the outputs. The red color coding represents thinner and the blue shows us the thicker regions. This shows us that some regions in these bubbles are more than twice as thick as others. And there are also more extreme cases where there is a six time difference between this and this part. You can see how the difference in thickness leads to waves of light interfering with the bubble and creating these beautiful rainbow patterns. You can't get this without a proper simulator like this one. Loving it. This variation in thicknesses is responsible for a selection of premium quality effects in a simulation. Beyond surface vortices, interference patterns can also be simulated deformation dependent rupturing of soap films. This incredible technique can simulate all of these phenomena. And now our big question. Ok, it simulates all of these, but how well does it do that? It is good enough to fool the human eye, but how does it compare to the strictest adversary of all, reality. I hope you know what's coming. Oh, yeah. Hold on to your papers, because now we will let reality be our judge and compare the simulated results to that. That is one of the biggest challenges in any kind of simulation research. So let's see. This is a piece of real footage of a curved soap film surface where these rainbow patterns get convected by an external force field. Beautiful. And now let's see the simulation. Wow. This has to be really close. Let's see them side by side and decide together. Whoa. The match in the swirly region here is just exceptional. Now, note that even if the algorithm is 100% correct, this experiment cannot be a perfect match because not only the physics of the soap film has to be simulated correctly, but the forces that move the rainbow patterns as well. We don't have this information from the real world footage, so the authors had to try to reproduce these forces, which is not part of the algorithm, but a property of the environment. So I would say that this footage is as close as one can possibly get. My goodness. Well done. So how much do we have to pay for this in terms of computation time? If you ask me, I would pay at the very least double for this. And now comes the best part. If you have been holding onto your paper so far, now squeeze that paper because in the cheaper cases, only 4 to 7% extra computation, which is outrageous. There is this more complex case with the large deforming sphere. In this case, the new technique indeed makes a huge difference. So how much extra computation do we have to pay for this? Only 31%. 31% extra competition for this. That is a fantastic deal. You can sign me up right away. As you see, the pace of progress in computer graphics research is absolutely incredible. And these simulations are just getting better and better by the day. Imagine what we will be able to do just two more papers down the line. What a time to be alive.
This episode has been supported by Lambda GPU Cloud. If you're looking for inexpensive cloud GPUs for AI, check out Lambda GPU Cloud. They've recently launched Quadro RTX 6000, RTX 8000 and V100 instances. And hold on to your papers because Lambda GPU Cloud can cost less than half of AWS and and Azure plus they are the only cloud service with 48GB RTX 8000. Join researchers at organizations like Apple, MIT and Caltech in using Lambda Cloud instances, workstations or servers. Make sure to go to lambdalabs.compapers to sign up for one of their amazing GPU instances today. Our thanks to Lambda for their long standing support and for helping us make.
Better videos for you. Thanks for watching and for your generous support and I'll see you next time.
