Dr. Zsolnai breaks down a 2019 graphics paper that speeds up viscous fluid simulations 4x without losing quality. The team tested it on honey dripping and a hot shower for a bunny, and Houdini adopted it within a year.
Dear fellow scholars, this is Two Minute Papers with Dr. Zsolnai. Feast your eyes upon this simulation from our previous episode that showcases a high.
Viscosity material that is honey.
The fact that honey is viscous means.
That it is a material that is highly resistant against deformation. In simpler words, if we can simulate viscosity, well, we can engage in the favorite pastimes of the computer graphics researcher. Or in other words, take some of these letters, throw them around, watch them slowly lose their previous shapes, and then, of course, destroy them in a spectacular manner. I love making simulations like this. And when I do, I want to see a lot of detail, which unfortunately means that I also have to run these simulations for a long time. So when I saw that this new technique promises to compute these about four times faster, it really grabbed my attention. Here is a visual demonstration of such a speed difference. Hmm. So four times you say? That sounds fantastic, but what's the catch here? This kind of speedup usually comes with cutting corners. So let's test the might of this new method through three examples of increasing difficulty. Experiment 1. Here is the regular simulation, and here is the new technique. So let's see the quality differences one more time. Well, I don't see any. Do you? Four times faster with no degradation in quality. Hmm. So far, so good. So now let's give it a harder example. Experiment number two. Varying viscosities and temperatures. In other words, let's give this bunny a hot shower. That was beautiful. And our question is again, how close is this to the slow reference simulation? Wow, this is really close. I have to look carefully to even have a fighting chance in finding a difference check mark. And I also wonder, can it deal with extremely detailed simulations? And now hold on to your papers for experiment number three. Dropping a viscous bunny on thin wires. And just look at the remnants of the poor bunny stuck in there. Loving it. Now, for almost every episode of this paper, I get comments. Caroj. This is all great, but when do I get to use this? When does this make it to the real world? These questions are completely justified. And. And the answer is right about now. You can use this right now. This paper was published in 2019 and now it appears to be already part of Houdini, one of the industry standard programs for visual effects and physics simulations. Tech transfer in just one year. Wow. Huge congratulations to Ryan Godade and his colleagues for this incredible paper. And huge respect to the folks at Houdini who keep outdoing themselves with these amazing updates.
This episode has been supported by weights.
And biases in this post. They show you how to monitor and optimize your GPU consumption during model training in real time with one line of code.
Weights and Biases provides tools to track your experiments in your deep learning projects. Their system is designed to save you.
A ton of time and money and.
It is actively used in projects at prestigious labs such as OpenAI, Toyota Research, GitHub, and more. And the best part is that Weights and Biases is free for all individuals, academics and open source projects. It really is as good as it gets. Make sure to Visit them through wnb.com papers or just click the link in the video description and you can get.
A free demo today.
I our thanks to Weights and Biases 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.
