Every year, you see us attend a conference you may or may not be familiar with called SIGGRAPH - an annual conference that fosters and celebrates innovation in Computer Graphics and interactive techniques. The event is about showcasing advancements and achievements amongst peers across the globe, a forum we have contributed to most years, this year being no different.
Despite the event being fully virtual this year, we have 5 contributions to the SIGGRAPH conference. What you may not know about Autodesk is that we don’t just develop software – we also contribute to the research community, welcoming other researchers to leverage our work, and kickstart a development cycle where we can in the end create something that helps our customers tell their stories.
So what’s on tap for Autodesk this year? We’ll be presenting two technical papers based on Arnold research, one technical talk based on Bifrost research, and we’ll be presenting in two courses: “Physically Based Shading” and “Advances in Monte-Carlo Rendering.”
Iliyan Georgiev from the Arnold team on Continuous Multiple Importance Sampling and Specular Manifold Sampling
In the Arnold team, we’re continuously researching complex rendering problems and regularly publish our results in various forums, including SIGGRAPH. We collaborate with several university labs across the globe, tackling open problems in rendering, and the result of these efforts this year are these technical papers.
The first one is titled "Continuous Multiple Importance Sampling" (in collaboration with The University of Tokyo), which presents a theoretical framework for efficiently combining stochastic sampling techniques. This can sound rather technical, and fortunately, the method works entirely behind the scenes to improve the efficiency of various rendering sub-problems and thus the overall performance.
The second paper is titled "Specular Manifold Sampling" and presents a technique for rendering complex caustics. Caustics are those beautiful illumination patterns we see when light is reflected or refracted at specular objects. Caustics can be seen in everyday life, for example, at the bottom of a swimming pool, or on a table due to light penetrating through a glass of wine. However, rendering such effects has been a notoriously difficult proposition for production renderers like Arnold, which is largely due to existing caustic solutions requiring pre-processing, caching, and often producing image artifacts.
Our paper provides an alternative solution that avoids pre-processing and artifacts, and it’s also designed to integrate well into production renderers.
Michael Nielsen, Morten Bojsen-Hansen, Robert Bridson, and Konstantinos Stamatelos from the Bifrost team on Auto-adaptivity
One of the innovations we’re contributing this year is related to smoke and fire simulation. This type of simulation for high-end VFX is very time-consuming and memory-intensive, and contrary to previous approaches, our work includes a novel workflow and algorithm for spatial adaptivity that makes spatial adaptivity for smoke simulation more automatic and less manual. We’ve named it Auto-Adaptivity.
Auto-adaptivity encompasses a new workflow and algorithm for spatially adaptive smoke and fire simulations which, based on a few input parameters, automatically combines any number of controls for adaptivity, such as - for example, where the smoke has high detail, where the flow rotates and so on -while automatically delivering and optimizing for the simulation quality.
For artists, it's important to be able to iterate faster on simulations to meet the vision of the supervisor or director, and the auto-adaptivity workflow accelerates this process.
The road to becoming an expert at running fluid simulations is notoriously convoluted, technical, difficult, and time-consuming, and having tools to optimize the performance of these simulations isn’t always worth the performance gains. Making performance-enhancing auto-adaptivity as straightforward as possible became a top priority for us, leading to the development of something that is as close to automatic as it comes.
This type of mindset can enrich the computer graphics world by embracing usability which in turn can lead to innovation and ultimately, improve the work-life of our users.
Niklas Harrysson and Iliyan Georgiev on MaterialX and Advances in Monte-Carlo Rendering
Finally, the SIGGRAPH courses program offers the chance for attendees to deepen their understanding of a particular topic and for experts to give back and share their knowledge. This year we're contributing to two courses.
The "Physically Based Shading" course presents recent ground-breaking work in topics like material modeling, sampling, volume rendering. Our contribution is a talk on MaterialX physically based shading nodes, which is a collaboration with our own Niklas Harrysson and Jonathan Stone from Lucasfilm.
The "Advances in Monte-Carlo Rendering" course presents the state-of-the-art methods in photo-realistic light transport simulation. Our contribution is a talk from Iliyan Georgiev on advanced sampling techniques for rendering complex lighting effects.