Program

Below is the program for Sunday, Monday, Tuesday and Wednesday. The alloted time for MAM presentations is 15 minutes plus 5 minutes for discussion and setup. EGSR presentations take 18 minutes plus 5 minutes for discussion and setup. There is additional information for presenters on the venue page.

Best Paper Award

The best paper award honors excellent work as well as a good presentation at the conference. The winner is selected by an anonymous jury of four members. Works in both the CGF Track and the EI&I track are eligible. This year, the winning paper is:

The authors received a Titan V that was generously donated by NVIDIA. Since it has been a close call, the jury decided to additionally name a runner-up for the award:

Congratulations to all of the authors for their outstanding work.

Keynotes

Unfortunately, Matt Pharr cannot attend the conference for personal reasons and his keynote has been canceled. Instead, there will be a keynote by Per Christensen.

Gambling in the Depths of High-Dimensional Spaces

Abstract:

Integration is a ubiquitous mathematical tool, and modern applications require integration across increasingly higher dimensional spaces. Unfortunately most of the intuitions that we take for granted in our low-dimensional, routine experiences don’t persist to these high-dimensional spaces which makes the development of scalable computational methodologies and algorithms all the more challenging. In this talk I will discuss the counterintuitive behavior of high-dimensional spaces and the consequences for statistical computation, in particular the unique advantages of Hamiltonian Monte Carlo.

Bio:

Michael Betancourt is the principal research scientist with Symplectomorphic, LLC where he develops theoretical and methodological tools to support practical Bayesian inference. He is also a core developer of Stan, where he implements and tests these tools. In addition to hosting tutorials and workshops on Bayesian inference with Stan he also collaborates on analyses in epidemiology, pharmacology, and physics, amongst others. Before moving into statistics, Michael earned a B.S. from the California Institute of Technology and a Ph.D. from the Massachusetts Institute of Technology, both in physics.

Interactive and Off-Line Path Tracing with RenderMan

Abstract:

RenderMan is a modern extensible and programmable path tracer with many features essential to handling the fiercely complex scenes in movie production. RenderMan has traditionally been focused on off-line rendering of high-quality final movie frames, but has recently been overhauled, targeting interactive rendering during modeling, texturing, lay-out, animation, and lighting. Path tracing has gone from being a pure research technique to now being the main rendering technique in many production renderers. In this talk Per Christensen will describe the use of path tracing for animated movies and visual effects, and will also describe advanced path tracing techniques such as bidirectional path tracing, progressive photon mapping, and vertex connection and merging (VCM). He will also touch upon current rendering projects at Pixar such as mixed CPU and GPU rendering and high-dimensional sample sequences specifically targeted at path tracing.

Bio:

Per Christensen is a principal software developer in Pixar's RenderMan group in Seattle. His main research interests are efficient ray tracing and global illumination in very complex scenes. He received an M.Sc. degree in electrical engineering from the Technical University of Denmark and a Ph.D. in computer science from the University of Washington. Prior to joining Pixar, he worked at ILM in San Rafael, Mental Images in Berlin, and Square USA in Honolulu. He has movie credits in Pixar movies since "Finding Nemo", and has received an Academy Award for his contributions to efficient point-based global illumination and ambient occlusion.

CANCELED: Offline and interactive rendering: Converging at last?

Abstract:

For decades, offline rendering and interactive rendering have been two largely separate worlds, where different constraints and requirements have led to the adoption of very different rendering techniques. Interactive rendering today is largely the realm of GPU-powered rasterization, where predictable performance and maintaining frame-rate are key requirements. In contrast, offline rendering has generally come to adopt CPU-based path tracing, given higher accuracy and realism expectations and a willingness to trade-off performance for flexibility. Recent work on the application of deep learning to image filtering and reconstruction in both realms points to the possibility that both types of rendering will converge to follow similar approaches: sparse sampling with path tracing followed by ML-driven reconstruction.
In this talk I will first discuss some of the constraints that both types of rendering face and how those have influenced them in the past. Next, I'll talk about some of the implications of incorporating ML into the image synthesis process as well as a few directions for further work in this area. Finally, I'll discuss ways in which the offline rendering community can contribute to interactive rendering, assuming this convergence continues.

Bio:

Matt Pharr is a distinguished research scientist at NVIDIA research, where he's working on ray tracing and applications of deep learning to rendering. He previously initiated and led the light field capture and rendering project in the VR group at Google, started Exluna (acquired by NVIDIA) and Neoptica (acquired by Intel), and wrote the ispc compiler. With Wenzel Jakob and Greg Humphreys, he is the author of Physically Based Rendering: From Theory to Implementation, for which he was awarded a Scientific and Technical Academy Award in 2014 for the influence the book has had on rendering for film production. He has B.S. in Computer Science from Yale and a Ph.D. from Stanford.