Schedule
Continental breakfast will be provided from the 20th to the 23rd, and a boxed lunch will be provided from the 21st to the 23rd.
A banquet dinner is planned at Le Castel on the 22nd, a few steps away from McGill.
Keynote Speakers
Liquid Content: an exploration of the future of culture and creativity
Imagine stories that shape-shift with every watch, creating a personal revolution for each viewer. Dive into the world of AI-powered storytelling that's turning culture on its head. In this talk, Matthieu will explore the concept of Liquid Content, where narratives break free from their confines, offering an endless canvas of creativity.
Matthieu Lorrain is a creative and technology pioneer in the fields of digital experiences & content innovation. He is currently Creative Lead, AI & Creativity Research at Google DeepMind. He is also the co-founder of fAke Artists, a creative collective exploring the future of post-reality experiences. Matthieu Lorrain has a long history working with global brands and tech companies to invent new types of user engagement. He has been exploring creative applications of emerging technologies for the last 20 years: ranging from interactive video to connected objects, augmented reality, and artificial intelligence. His most recent work focuses on how generative AI can supercharge the creative experience. His past projects have received multiple accolades from global organizations including Emmy Awards, Cannes Lions (Gold), Clios (Gold), Webby's, Tribeca Film Festival, '#1 Product Hunt of the Day' and FWA. Matthieu is a guest lecturer at Columbia University, where he delivered the inaugural masterclass on AI & Filmmaking in 2024. He is also frequently invited as featured speaker at major conferences. He has previously spoken at Cannes Lions (3x), SXSW, Spike Asia, 4A's Createtech and the NYC Tech Forum. Born & raised in the French Alps, Matthieu has lived in Rio de Janeiro, Montreal and Paris before moving to New York City in 2011. He holds a Master's degree in Cultural Studies from Institut d'Etudes Politiques and another in Marketing & Communication from ESCP Paris Business School.
Is data the only lever for designing interactive simulations?
The design of interactive simulations has always been struggling on the trade-off between accuracy and computation time performances. These last years, the rise of data-driven approaches has paved the way for new models offering outstanding results for several use cases. Thus, if the use of data is nowadays commonly accepted for some scenarios, it often remains uncertain how, when or where data can outperform more conventional approaches for designing interactive simulations. In this talk, I will illustrate how we can combine data-driven and model-based approaches for designing interactive simulations within the context of robotics and virtual reality applications.
Maud Marchal is a Full Professor in Computer Science at Univ. Rennes, INSA/IRISA in France. She is also a Junior Member of Institut Universitaire de France since 2018. She works on physics-based simulation since her PhD in 2006 at University Joseph Fourier, Grenoble. Since 2008 and her position at INSA, she has explored and contributed to novel Virtual Reality and robotics applications, gathering her expertise on haptic feedback, 3D interaction techniques and interactive physics-based simulations. She is involved in program committees of major conferences of computer graphics, virtual reality and haptics and Associate Editor of IEEE Transactions on Visualization and Computer Graphics, IEEE Transactions on Haptics, ACM Transactions on Applied Perception and Computers & Graphics. She has notably been Program Chair of IEEE Virtual Reality Conference in 2018, 2020 and 2021, Program Chair of IEEE Symposium on Mixed and Augmented Reality in 2021 and 2023 and General Chair of ACM SIGGRAPH/Eurographics Symposium on Computer Animation in 2018 and Eurohaptics in 2024.
Expressive Facial Modeling and Animation
Humans are hard-wired to see and interpret minute facial detail. The rich signals we extract from facial expressions set high expectations for computer-generated facial imagery. This talk focuses on the science and art of expressive facial animation. Specifically, aspects of facial anatomy, biomechanics, linguistics and perceptual psychology will be used to motivate and describe the construction of geometric face rigs, and techniques for the animator-centric creation of emotion, expression and speech animation from input images, audio and video. In some measure the talk will reveal some of the technological innovations that enabled the design and creation of faces in games like Cyberpunk 2077 (Game of the year 2020), and films like Avatar: the way of water (Best VFX Oscar 2023).
Karan Singh is a Professor of Computer Science at the University of Toronto. His research interests lie at the intersection of art, Computer Graphics (CG) and Human Computer Interaction (HCI): spanning interactive modeling and animation, visual perception, visualization and Augmented/Virtual Reality. Karan has been a research and development lead on the technical Oscar (2003) winning modeling and animation system Maya. He has co-founded multiple companies, most recently JALI Research. He was the R&D Director for the 2005 Oscar winning animated short film Ryan. His recent research in facial animation has been used on characters in AAA games like Cyberpunk 2077 and Call of Duty: Modern Warfare 2, and films like Avatar: the way of water (Best VFX Oscar 2023).
Graduate Summer School
Introduction to Optimization for Simulation
Energy minimization is central in physics-based simulation and geometry processing, with applications ranging from mesh parameterization to elastodynamic simulation. This talk will cover both classic techniques and the most recent advances for elastic energy minimization, as well as their trade-offs. We will begin with the classic gradient descent and Newton’s method. Then we will dive into more advanced optimization techniques, exploring various ways to inject second-order information into the optimization while maintaining low computational cost. The course will also address practical aspects of implementing these algorithms, highlighting common challenges and best practices. Attendees will learn how to choose the optimal optimization strategy based on their specific problem and application requirements, balancing computational efficiency, accuracy, and robustness.
Honglin Chen is a third-year PhD student in Computer Science at Columbia University, advised by Changxi Zheng. Her research lies in the intersection of physics-based simulation, geometric processing, numerical optimization, and (a little bit of) machine learning. She obtained her MSc from University of Toronto advised by David I.W. Levin, and her B.Eng. from Zhejiang University. Currently, her research focuses on developing easy-to-use simulation tools and numerical methods for physics-based animation and real-world creative tasks. She has interned at Adobe, Meta, Nvidia and Microsoft Research Asia during her studies.
Incompressible fluids in VFX
In this talk we will go over the main components needed to build an incompressible fluid solver. This will include common techniques such as operator splitting, a derivation of pressure projection, spatial- and temporal discretization, and semi-Lagrangian advection. In the end I will show examples of how on can use these tools in the VFX industry to produce everything from waterfalls to breaking waves. In particular I will highlight a useful coupling strategy between incompressible fluids and solids that can simulate both capes fluttering in the wind as well as bubbles rising in the ocean.
I am Joel Wretborn, a Simulation researcher at Wētā FX and PhD student at the University of Waterloo under supervision of Christopher Batty. My main area of interest is the animation of fluids, and in particular multi-phase media where a fluid couples with some other material. My work has been show-cased in movies such as Avengers: Endgame and Kingdom of the Planet of the Apes, and we were recently awarded the VES Emerging Technology Award for our work on fluids for Avatar 2: The Way of Water.
Introduction to Elastodynamic Simulation for Solids and Cloth
The simulation of deformable elastic objects has become a popular tool in visual effects, games, and interactive virtual environments. In this course, students will learn continuous mathematics that describe the motion of physical objects, explore discretization techniques, and solve the resulting discrete equations efficiently and robustly. The course begins with a discussion on a simple mass-spring system to introduce the optimization time integration framework. It then briefly covers the finite element method, widely used for its versatility in representing elastic solids and cloth with intricate geometric features and diverse material properties. We will also demonstrate how recent advancements in elastodynamic simulation are rooted in this shared framework, which supports extensions such as subspace methods for fast simulations, enhancements to rig-based animations with physical secondary motion, and the integration of multilevel methods for rapid previews and enhanced user interactivity, among other applications.
Jiayi (Eris) Zhang is a 3rd year PhD candidate at Stanford University, advised by Prof. Doug James. She completed her undergraduate studies in computer science and math at the University of Toronto, advised by Prof. Alec Jacobson. She works on physics-based animation, geometry processing and numerical optimization. Currently, her research primarily focuses on developing intelligent algorithms, models and tools for enhancing user creativity and productivity in design, animation and simulation. She has also been interning at Adobe for a few summers, working closely with Dr. Danny Kaufman.
Introduction to Physics-based Character Control
In this lecture, I will dive into the field of physics-based character control, an approach that leverages physical simulations to create lifelike and responsive motions for digital characters. After a quick introduction to the three fundamental components for character control: physics-simulation, characters and motion controllers, the rest of the lecture is spent on learning motion controllers. Students will learn what are the most recent techniques to create a motion controller using reinforcement and imitation learning. The lecture will cover different reward structures, ways to use motion capture data to aid the learning, and more advanced techniques to learn harder and more complex motions. Finally, we will move to showcasing different examples with humanoid motion control, virtual characters, and robots in the real world.
Daniele Reda is a 5th year PhD candidate at University of British Columbia, working on physics-based character control, reinforcement learning, planning and robotics with professor Michiel van de Panne. His interests lie in making characters and robots move and teach them to do so autonomously while solving hard problems. He has also spent time at Sanctuary AI teaching humanoid robots how to use their hands to grasp different objects; at Meta Reality Labs to make animated characters move in virtual worlds; at Inverted AI on realistic driving simulators, and at Wayve, teaching cars how to drive in traffic.
Doctoral dissertation award
Conquering Contact: Provably Robust Simulation with Contact
Physical simulation has become a critical component of many scientific and engineering areas including computer graphics, robotics, biomechanics, and computational physics. However, the simulation of frictional contact is a challenging problem with many parameters to carefully adjust, and many existing methods allow for interpenetration of objects, which can lead to a variety of issues, including non-physical behavior, numerical instability, and even simulation failure. To address these issues, we introduced the Incremental Potential Contact (IPC) method in 2020, which is the first simulation algorithm for deformable and rigid bodies that is unconditionally robust, requires minimal parameter tuning, and provides a direct way of controlling the trade-off between running time and accuracy.
In this talk, I will discuss the work we have done since to provide provable guarantees of robustness and improved accuracy. Namely, we introduce a large-scale benchmark of continuous collision detection (CCD) algorithms (a core component of the IPC algorithm), and we introduce the first efficient CCD algorithm that is provably conservative. For improved accuracy and efficiency, I will discuss how we can utilize nonlinear finite element analysis and physically-based adaptive meshing within the IPC framework. Additionally, I will discuss the benefits of using IPC as a differentiable simulator, opening the doors to a variety of applications, including inverse design, parameter acquisition, and bridging the sim-to-real gap for challenging contact scenarios. Lastly, I will discuss our open-source software efforts, including the IPC Toolkit and PolyFEM, aimed towards democratizing these techniques for the broader community.
The SCA 2024 best dissertation award goes to Dr. Zachary Ferguson to recognize his contributions to automatic simulations with contact that support complex geometric scenes as well as its applications in shape design, biomechanics and soft robotics.
Zachary Ferguson earned his Ph.D. in Computer Science from New York University in May 2023, under the supervision of Daniele Panozzo. As a member of the Geometric Computing Lab, he conducted research in computer graphics, physical simulation, and digital fabrication. His work focused on designing robust algorithms that minimize interactive user input, enabling the automatic simulation of numerous complex designs and scenarios. Following his Ph.D., Zachary was a Postdoctoral Associate at MIT's Computer Science and Artificial Intelligence Laboratory, where he specialized in physical simulation and metamaterial design. He is currently a Research Scientist at CLO Virtual Fashion, where he works on physically-based animation, simulation, and geometry processing. Zachary is an active participant in the research community. He is serving as the poster chair for this year's ACM Motion, Interaction, and Games (MIG) conference and regularly reviews papers for journals and conferences such as SIGGRAPH, SIGGRAPH Asia, Eurographics, and Computer & Graphics. His research achievements have been recognized with several prestigious awards, including the ACM SIGGRAPH Outstanding Doctoral Dissertation Award (2024), the Adobe Research Fellowship (2022), the Dean's Dissertation Fellowship (2022-2023), and the Jacob T. Schwartz Ph.D. Fellowship (2021).
Early Career Researcher Award
Advancing Physics-based Animation with Adaptive Discretization and Machine Learning
Efficiency, robustness, and fidelity are critical features of physics-based animation (PBA) techniques. However, achieving optimal performance across all these aspects remains a significant challenge, limiting the scope of existing methods. Optimization time integration is a promising framework that can ensure robustness and fidelity when solved to sufficient accuracy and offers opportunities for efficiency improvements based on numerical methods. In this talk, I will present our recent advancements in PBA through adaptive discretization. This includes a contact proxy time-splitting method for penetration-free simulation of solid-fluid interactions, and a multi-resolution approach for efficient, high-quality simulation of thick garments. Furthermore, I will discuss leveraging machine learning to develop surrogate models for accelerated simulations in specific scenarios.
The SCA 2024 Early Career Researcher Award is presented to Dr. Minchen Li in recognition of his contributions to enhancing the robustness, efficiency, and fidelity of physics-based animation techniques. His work on adaptive discretization and machine learning has significantly advanced the field, with broad applications in computational mechanics, virtual reality, and robotics.
Minchen Li is an Assistant Professor in the Computer Science Department at Carnegie Mellon University, a position he has held since September 2023. Prior to this, he was an Assistant Adjunct Professor in the Department of Mathematics at UCLA, working in the AIVC Lab. He earned his Ph.D. in 2020 from the SIG Center for Computer Graphics at the University of Pennsylvania, under the guidance of Chenfanfu Jiang. Minchen's research accomplishments have been recognized with several prestigious awards, including the ACM SIGGRAPH Outstanding Doctoral Dissertation Award (2021), the Symposium on Computer Animation Doctoral Dissertation Award (2021), the Adobe Research Fellowship (2020), and the Mitacs Globalink Graduate Fellowship (2015-2016). He is an active member of the research community, regularly serving as a program committee member for conferences such as ACM SIGGRAPH, Eurographics, SCA, and CGI, as well as an external reviewer for journals including IEEE TVCG, IEEE ICRA, and ACM TOG.
Words of Wisdom Session
This year we revisit the Words of Wisdom (WoW) session that took place at SCA 2015 in Los Angeles.
This session invites computer animation experts to share ideas or thoughts to the community. If you have a scientific or even philosophical message to deliver which is personal, insightful, generous, original (and possibly provocative, but always benevolent and scientifically argumented), then we want to hear from you! No format nor topic is imposed, except that your message should be somehow related to computer animation, it should deal with science and techniques, it should be respectful to others and constructive enough to be useful and inspiring for many people (students and/or faculty and/or people in industry). It does not need to be broad and can be focused on a specific point. It could, for instance, fall into one of the following categories:
- An important challenge to be taken in computer animation
- The story of a failure (and what to learn from it)
- A research topic (never found time to work on), generously offered to the community
- A methodology/research direction/trend of our field you wish to celebrate/criticize/improve
- An analysis of the positioning of Computer Animation with respect to other fields
- ...
Confirmed speakers:
- Daniel Holden, Epic Games
- Julien Pettre, Inria
- Marie-Paule Cani, École Polytechnique
- Victor Zordan, Roblox
Paper Sessions
Character Animation I: Synthesis and Capture
Chaired by Sheldon Andrews
Pose-to-Motion: Cross-Domain Motion Retargeting with Pose Prior
Qingqing Zhao, Peizhuo Li, Wang Yifan, Olga Sorkine-Hornung, Gordon Wetzstein
Long-term Motion In-betweening via Keyframe Prediction
Seokhyeon Hong, Haemin Kim, Kyungmin Cho, Junyong Noh
ADAPT: AI-Driven Artefact Purging Technique for IMU Based Motion Capture
Paul Schreiner, Rasmus Netterstrøm, Hang Yin, Sune Darkner, Kenny Erleben
Diffusion-based Human Motion Style Transfer with Semantic Guidance
Lei Hu, Zihao Zhang, Yongjing Ye, Yiwen Xu, Shihong Xia
Physics I: Fluids, Shells and Natural Phenomena
Chaired by Alexandre Mercier-Aubin
Multiphase Viscoelastic Non-Newtonian Fluid Simulation
Yalan Zhang, Shen Long, Yanrui Xu, Xiaokun Wang, Chao Yao, Jiri Kosinka, Steffen Frey, Alexandru Telea, Xiaojuan Ban
Reconstruction of implicit surfaces from fluid particles using convolutional neural networks
Chen Zhao, Tamar Shinar, Craig Schroeder
Unerosion: Simulating Terrain Evolution Back in Time
Zhanyu Yang, Guillaume Cordonnier, Marie-Paule Cani, Christian Perrenoud, Bedrich Benes
Curved Three-Director Cosserat Shells with Strong Coupling
Fabian Löschner, José Antonio Fernández-Fernández, Stefan Rhys Jeske, Jan Bender
Character Animation II: Control
Chaired by Julien Pettré
Learning to Move Like Professional Counter-Strike Players
David Durst, Feng Xie, Vishnu Sarukkai, Brennan Shacklett, Iuri Frosio, Chen Tessler, Joohwan Kim, Carly Taylor, Gilbert Bernstein, Sanjiban Choudhury, Pat Hanrahan, Kayvon Fatahalian
PartwiseMPC: Interactive Control of Contact-Guided Motions
Niloofar Khoshsiyar, Ruiyu Gou, Tianhong Zhou, Sheldon Andrews, Michiel van de Panne
VMP: Versatile Motion Priors for Robustly Tracking Motion on Physical Characters
Agon Serifi, Ruben Grandia, Espen Knoop, Markus Gross, Moritz Bächer
Physics II: Cutting & Colliding
Chaired by Minchen Li
Generalized eXtended Finite Element Method for Deformable Cutting via Boolean Operations
Quoc-Minh Ton-That, Paul Kry, Sheldon Andrews
Strongly Coupled Simulation of Magnetic Rigid Bodies
Lukas Westhofen, José Antonio Fernández-Fernández, Stefan Rhys Jeske, Jan Bender
A Multi-Layer Solver for XPBD
Alexandre Mercier-Aubin, Paul Kry
Robust and Artefact-Free Deformable Contact with Smooth Surface Representations
Yinwei Du, Yue Li, Stelian Coros, Bernhard Thomaszewski
Animation Tools
Chaired by Daniel Holden
Creating a 3D Mesh in A-pose from a Single Image for Character Rigging
Seunghwan Lee, C. Karen Liu
SketchAnim: Real-time sketch animation transfer from videos
Gaurav Rai, Shreyas Gupta, Ojaswa Sharma
Garment Animation NeRF with Color Editing
Renke Wang, Meng Zhang, Jun Li, Jian Yang
Learning to Play Guitar with Robotic Hands
Chaoyi Luo, Pengbin Tang, Yuqi Ma, Dongjin Huang
Gesture and Gaze Animation, and Cinematography
Chaired by Anne-Hélène Olivier
LLAniMAtion: LLAMA Driven Gesture Animation
Jonathan Windle, Iain Matthews, Sarah Taylor
Real-time multi-map saliency-driven gaze behavior for non-conversational characters
Ific Goudé, Alexandre Bruckert, Anne-Hélène Olivier, Julien Pettré, Rémi Cozot, Kadi Bouatouch, Marc Christie, Ludovic Hoyet
Reactive Gaze during Locomotion in Natural Environments
Julia Melgaré, Damien Rohmer, Soraia Musse, Marie-Paule Cani
Generating Flight Summaries Conforming to Cinematographic Principles
Christophe Lino, Marie-Paule Cani
Posters
1 - Smoothed-Hinge Model for Cloth Simulation
Qixin Liang
2 - Markerless Multi-view Multi-person Tracking for Combat Sports
Hossein Feiz, David Labbé, Sheldon Andrews
3 - Brittle Fracture Animation with VQ-VAE-Based Generative Method
Yuhang Huang, Takashi Kanai
4 - Organic Brushstrokes
William J. Joel
5 - Learning Climbing Controllers for Physics-Based Characters
Kyungwon Kang, Taehong Gu, Taesoo Kwon
6 - Neural Implicit Reduced Fluid Simulation
Yuanyuan Tao, Ivan Puhachov, Derek Nowrouzezahrai, Paul Kry
7 - A Differentiable Material Point Method Framework for Shape Morphing
Michael Xu, Changyong Song, David Levin, David Hyde
8 - Adaptive Sampling for Simulating Granular Materials
Samraat Gupta, John Keyser
9 - Art-directable expressive oscillation behavior for rigged characters
Karim Salem, Damien Rohmer, Niranjan Kalyanasundaram, Victor Zordan
10 - Rigid Body Adversarial Attacks
Aravind Ramakrishnan, David Levin, Alec Jacobson
11 - Data-driven Friction for Real-time Applications
Loic Nassif, Omar Zoubir, Sheldon Andrews, Paul Kry
12 - Fast simulation of viscous lava flow using Green's functions as a smoothing kernel
Yannis Kedadry, Guillaume Cordonnier