Dr. KƔroly Zsolnai-FehƩr covers a new method for simulating thin-film fluids like soap bubbles. The research shows how color shifts as the film thins before it bursts, and runs at seconds per frame on a laptop. Four demos include oscillating bubbles, Rayleigh-Taylor instability, and a catenoid surface collapsing under tension.
Dear fellow scholars, this is Two Minute Papers with Dr. Zsolnai. Today we are going to simulate these absolutely beautiful thin film structures. You see, computer graphics researchers have been writing physics simulation programs for decades now, and the pace of progress in this research area is absolutely stunning. Here are three examples of where we are at the moment. One, this work was able to create a breathtaking honeycoiling simulation. I find it absolutely amazing that through the power of computer graphics research, all this was possible four years ago. And the realistic simulation works just kept coming in. This work appeared just one year ago and could simulate not only a piece of viscous fluid, but also deal with glagging and coalescing bubbles. And three, this particular one is blazing fast. So much so that it can simulate this damn break scene in about five frames per second, not seconds per frame, while it can run this water drop scene with about seven frames per second. Remember, this simulates quantities like velocity, pressure and more for several million particles. This quickly. Very impressive. So are we done here? Is there anything else left to be done in fluid simulation research? Well, hold on to your papers and check this out. This new paper can simulate thin film phenomena. What does that mean? Well, four things. First, here is a beautiful oscillating soap bubble. Yes, its color varies as a function of the evolving film thickness. But that's not all. Let's poke it. And then did you see that? It can even simulate it bursting into tiny sparkly droplets. Phew. One more time. Loving it. Second, it can simulate one of my favorites, the Rayleigh Taylor instability. The upper half of the thin film has a larger density, while the lower half carries a larger volume. Essentially, this is the phenomenon when two fluids of different densities meet. And what is the result? Turbulence. First, the interface between the two is well defined, but over time, it slowly disintegrates into this beautiful straight swirly pattern. Oh yeah, look. And it just keeps on going and going. Third, Ah, yes, the catenoid experiment. What is that? This is a surface tension driven deformation experiment where the film is trying to shrink as we move the two rims away from each other, forming this catenoid surface. Of course, we won't stop there. What happens when we keep moving them away? What do you think? Please stop the video and let me know in the comments below. I'll wait a little. Thank you. Now then, the membrane keeps shrinking until, yes, it finally collapses into a small droplet. The authors also went the extra mile and did the most difficult thing for any physics simulation paper, comparing the results to reality. So is this just Good enough to fool the untrained human eye, or is this the real deal? Well, look at this. This is an actual photograph of the Catanoid experiment. And this is the simulation. Dear fellow scholars, that is a clean simulation right there. And fourth, a thin film within a square is subjected to a gravitational pull that is changing over time and the result is more swirly patterns. So how quickly can we perform all this? Disregard the FPS. This is the inverse of the time step size and is mainly information for fellow researchers. For now, gaze upon the time per frame column. And my goodness, this is blazing fast too. It takes less than a second per frame for the Catanoid experiment. This is one of the cheaper ones. And all this on a laptop. Wow. Now the most expensive experiment in this paper was the Rayleigh Taylor instability. This took about 13 seconds per frame. This is not bad at all. We can get a proper simulation of this quality within an hour or so. However, note that the authors also used a big honking machine to compute this scene. And remember, this paper is not about optimization, but it is about making the impossible possible and it is doing all that swiftly. Huge congratulations to the authors. What a time to be alive. Perceptilabs is a Visual API for TensorFlow, carefully designed to make machine learning as intuitive as possible. This gives you a faster way to build out models with more transparency into how your model is architected, how it performs, and how to debug it. Look, it lets you toggle between the visual modeler and the code editor. It even generates visualizations for all the model variables and gives you recommendations both during modeling and training, and does all this automatically. I only wish I had a tool like this when I was working on my neural networks during my PhD years. Visit perceptilabs.com papers to easily install the free local version of their system today. Our thanks to Perceptilabs for their 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.
