Siggraph Highlights
A Deep Dive into Universal Scene Description and Hydra
Universal Scene Description has enjoyed a huge uptick in support, due to integration in DCC apps such as Houdini, and Apple building the RealityKit pipeline around it.
Hydra offers a render abstraction that allows evaluating many renderers side by side versus identical scene descriptions.
https://graphics.pixar.com/usd/docs/index.html
Advances in Real-Time Rendering Part I & II
Syllabus available:
http://advances.realtimerendering.com/s2019/index.htm
Highlights:
- A Journey Through Implementing Multiscattering BRDFs and Area Lights
- Strand-based Hair Rendering in Frostbite
- Interactive Wind and Vegetation in 'God Of War'
- Multi-resolution Ocean Rendering in Crest Ocean System
- Creating the Atmospheric World of Red Dead Redemption 2: A Complete and Integrated Solution
Takeaway:
Advanced real time rendering uses ray tracing in one way or another as part of the final image composition.
Academy Software Foundation Open Source Day
The Linux Foundation and the Academy of Motion Pictures have started an effort to collaborate on open source development. Several previously moribund, yet still highly in-use libraries, such as OpenEXR, OpenColorIO, and others are now developed under the aegis of the ASWF. This has renewed interest and participation of the community.
The day also highlighted other upcoming open source projects of which several are likely to be of use to us, such as MaterialX.
Real-Time Live
There was much repetition of material this year, with incremental improvements, rather than anything genuinely new. One exception was Hao Li's presentation on Hair Grooming, where an innovative work flow was presented. A user draws guide strands of hair, and a neural network selects plausible hair stylings from a learned database.
http://www.hao-li.com/publications/papers/uist2019HBIDDHM.pdf
Learning To Move
This session was great to see.
Synthesis of Biologically Realistic Human Motion Using Joint Torque Actuation
Yifeng Jiang (Georgia Institute of Technology), Tom Wouwe, Friedl De Groote (KU Leuven), C. Karen Liu (Georgia Institute of Technology)
Scalable Muscle-actuated Human Simulation and Control
Seunghwan Lee (Seoul National University), Kyoungmin Lee, Moonseok Park (Seoul National University Bundang Hosiptial), Jehee Lee (Seoul National University)
Fluids I, Silly Rubber, Fluids II
Simulation technology continues to advance, in particular, approximations and improvements in system coupling are a new frontier.
Interlinked SPH Pressure Solvers for Strong Fluid-Rigid Coupling (TOG Paper) Christoph Gissler (University of Freiburg and FIFTY2 Technology), Andreas Peer, Stefan Band (University of Freiburg), Jan Bender (RWTH Aachen), Matthias Teschner (University of Freiburg)
An Adaptive Variational Finite Difference Framework for Efficient Symmetric Octree Viscosity
Ryan Goldade, Yipeng Wang (University of Waterloo), Mridul Aanjaneya (Rutgers University), Christopher Batty (University of Waterloo)
Silly Rubber: An Implicit Material Point Method for Simulating Non-equilibrated Viscoelastic and Elastoplastic Solids
Yu Fang, Minchen Li, Ming Gao, Chenfanfu Jiang (University of Pennsylvania)
Implicit Untangling: A robust solution for modeling layered clothing
Thomas Buffet, Damien Rohmer (Ecole Polytechnique), Loic Barthe (IRIT), Laurence Boissieux (INRIA), Marie-Paule Cani (Ecole Polytechnique)
Efficient and Conservative Fluids Using Bidirectional Mapping
Ziyin Qu* (AICFVE and University of Pennsylvania), Xinxin Zhang* (AICFVE), Ming Gao, Chenfanfu Jiang (University of Pennsylvania), Baoquan Chen (Peking University) (*joint 1st authors)
On Bubble Rings and Ink Chandeliers
Marcel Padilla, Albert Chern, Felix Knoppel, Ulrich Pinkall (Technische Universitat Berlin), Peter Schroder (Caltech)
Fundamental solutions for water wave animation
Camille Schreck, Christian Hafner, Chris Wojtan (IST Austria)
Animation and Skinning
I found direct delta mush skinning to be particularly disappointing, it seems yet another ad hoc deformation model with more cost than LBS, that simply moves errors and approximations around, with no true convergence on any sort of improvement over an artist and a good set of LBS authoring materials.
Direct Delta Mush Skinning and Variants
Binh Huy Le (SEED - Electronic Arts), JP Lewis (Google AI)
NeuroSkinning
This was very clever. Given a large corpus of artist authored character assets, a neural network proposes skin weights for new character assets. The key trick here is that a super-skeleton is derived, represented as a sparse matrix, which allows a neural network to generalize bindings across character models.
NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks
Lijuan Liu, Youyi Zheng, Di Tang, Yi Yuan, Changjie Fan, Kun Zhou (Zhejiang University)