Distant Collision Response in Rigid Body Simulations

We use a finite element model to predict the vibration response of objects in a rigid body simulation, such that rigid objects are augmented to provide a plausible elastic collision response between distant objects due to vibration. We start with a generalized eigenvalue decomposition of the elastic model to precompute a response to an impact at any point on an elastic object with fixed boundary conditions. Then, given a collision between objects, we generate an approximate response impulse to distribute to other objects already in contact with the colliding bodies. This can lead to distant impacts causing an object to slip, or a delicate stack of objects to fall. We also use a geodesic distance based spatial attenuation approximation for travelling waves in objects to respond to an impact at one contact with an impulse at other locations. This response ultimately allows a long distance relationship between contacts, both across a single object being struck, but also traversing the contact graph of a larger collection of objects. We qualitatively validate our approach with a ground truth simulation, and demonstrate a number of scenarios where a long distance relationship between contacts is valuable.     

Fully Convolutional Graph Neural Networks for Parametric Virtual Try-On

We present a learning-based approach for virtual try-on applications based on a fully convolutional graph neural network. In contrast to existing data-driven models, which are trained for a specific garment or mesh topology, our fully convolutional model can cope with a large family of garments, represented as parametric predefined 2D panels with arbitrary mesh topology, including long dresses, shirts, and tight tops. Under the hood, our novel geometric deep learning approach learns to drape 3D garments by decoupling the three different sources of deformations that condition the fit of clothing: garment type, target body shape, and material. Specifically, we first learn a regressor that predicts the 3D drape of the input parametric garment when worn by a mean body shape. Then, after a mesh topology optimization step where we generate a sufficient level of detail for the input garment type, we further deform the mesh to reproduce deformations caused by the target body shape. Finally, we predict fine-scale details such as wrinkles that depend mostly on the garment material. We qualitatively and quantitatively demonstrate that our fully convolutional approach outperforms existing methods in terms of generalization capabilities and memory requirements, and therefore it opens the door to more general learning-based models for virtual try-on applications.     

Cloth and Skin Deformation with a Triangle Mesh Based Convolutional Neural Network

We introduce a triangle mesh based convolutional neural network. The proposed network structure can be used for problems where input and/or output are defined on a manifold triangle mesh with or without boundary. We demonstrate its applications in cloth upsampling, adding back details to Principal Component Analysis (PCA) compressed cloth, regressing clothing deformation from character poses, and regressing hand skin deformation from bones' joint angles. The data used for training in this work are generated from high resolution extended position based dynamics (XPBD) physics simulations with small time steps and high iteration counts and from an offline FEM simulator, but it can come from other sources. The inference time of our prototype implementation, depending on the mesh resolution and the network size, can provide between 4 to 134 times faster than a GPU based simulator. The inference also only needs to be done for meshes currently visible by the camera.     

Latent Space Subdivision: Stable and Controllable Time Predictions for Fluid Flow

We propose an end-to-end trained neural network architecture to robustly predict the complex dynamics of fluid flows with high temporal stability. We focus on single-phase smoke simulations in 2D and 3D based on the incompressible Navier-Stokes (NS) equations, which are relevant for a wide range of practical problems. To achieve stable predictions for long-term flow sequences with linear execution times, a convolutional neural network (CNN) is trained for spatial compression in combination with a temporal prediction network that consists of stacked Long Short-Term Memory (LSTM) layers. Our core contribution is a novel latent space subdivision (LSS) to separate the respective input quantities into individual parts of the encoded latent space domain. As a result, this allows to distinctively alter the encoded quantities without interfering with the remaining latent space values and hence maximizes external control. By selectively overwriting parts of the predicted latent space points, our proposed method is capable to robustly predict long-term sequences of complex physics problems, like the flow of fluids. In addition, we highlight the benefits of a recurrent training on the latent space creation, which is performed by the spatial compression network. Furthermore, we thoroughly evaluate and discuss several different components of our method.     

ALLSTEPS: Curriculum-driven Learning of Stepping Stone Skills

Humans are highly adept at walking in environments with foot placement constraints, including stepping-stone scenarios where footstep locations are fully constrained. Finding good solutions to stepping-stone locomotion is a longstanding and fundamental challenge for animation and robotics. We present fully learned solutions to this difficult problem using reinforcement learning. We demonstrate the importance of a curriculum for efficient learning and evaluate four possible curriculum choices compared to a non-curriculum baseline. Results are presented for a simulated humanoid, a realistic bipedal robot simulation and a monster character, in each case producing robust, plausible motions for challenging stepping stone sequences and terrains.     

Probabilistic Character Motion Synthesis using a Hierarchical Deep Latent Variable Model

We present a probabilistic framework to generate character animations based on weak control signals, such that the synthesized motions are realistic while retaining the stochastic nature of human movement. The proposed architecture, which is designed as a hierarchical recurrent model, maps each sub-sequence of motions into a stochastic latent code using a variational autoencoder extended over the temporal domain. We also propose an objective function which respects the impact of each joint on the pose and compares the joint angles based on angular distance. We use two novel quantitative protocols and human qualitative assessment to demonstrate the ability of our model to generate convincing and diverse periodic and non-periodic motion sequences without the need for strong control signals.     

Non-linear Rough 2D Animation using Transient Embeddings

Traditional 2D animation requires time and dedication since tens of thousands of frames need to be drawn by hand for a typical production. Many computer-assisted methods have been proposed to automatize the generation of inbetween frames from a set of clean line drawings, but they are all limited by a rigid workflow and a lack of artistic controls, which is in the most part due to the one-to-one stroke matching and interpolation problems they attempt to solve. In this work, we take a novel view on those problems by focusing on an earlier phase of the animation process that uses rough drawings (i.e., sketches). Our key idea is to recast the matching and interpolation problems so that they apply to transient embeddings, which are groups of strokes that only exist for a few keyframes. A transient embedding carries strokes between keyframes both forward and backward in time through a sequence of transformed lattices. Forward and backward strokes are then cross-faded using their thickness to yield rough inbetweens. With our approach, complex topological changes may be introduced while preserving visual motion continuity. As demonstrated on state-of-the-art 2D animation exercises, our system provides unprecedented artistic control through the non-linear exploration of movements and dynamics in real-time.     

Physics-Informed Neural Corrector for Deformation-based Fluid Control

Controlling fluid simulations is notoriously difficult due to its high computational cost and the fact that user control inputs can cause unphysical motion. We present an interactive method for deformation-based fluid control. Our method aims at balancing the direct deformations of fluid fields and the preservation of physical characteristics. We train convolutional neural networks with physics-inspired loss functions together with a differentiable fluid simulator, and provide an efficient workflow for flow manipulations at test time. We demonstrate diverse test cases to analyze our carefully designed objectives and show that they lead to physical and eventually visually appealing modifications on edited fluid data.     

An Optimization-based SPH Solver for Simulation of Hyperelastic Solids

This paper proposes a novel method for simulating hyperelastic solids with Smoothed Particle Hydrodynamics (SPH). The proposed method extends the coverage of the state-of-the-art elastic SPH solid method to include different types of hyperelastic materials, such as the Neo-Hookean and the St. Venant-Kirchoff models. To this end, we reformulate an implicit integration scheme for SPH elastic solids into an optimization problem and solve the problem using a general-purpose quasi-Newton method. Our experiments show that the Limited-memory BFGS (L-BFGS) algorithm can be employed to efficiently solve our optimization problem in the SPH framework and demonstrate its stable and efficient simulations for complex materials in the SPH framework. Thanks to the nature of our unified representation for both solids and fluids, the SPH formulation simplifies coupling between different materials and handling collisions.     

Real Time Multiscale Rendering of Dense Dynamic Stackings

Dense dynamic aggregates of similar elements are frequent in natural phenomena and challenging to render under full real time constraints. The optimal representation to render them changes drastically depending on the distance at which they are observed, ranging from sets of detailed textured meshes for near views to point clouds for distant ones. Our multiscale representation use impostors to achieve the mid-range transition from mesh-based to point-based scales. To ensure a visual continuum, the impostor model should match as closely as possible the mesh on one side, and reduce to a single pixel response that equals point rendering on the other. In this paper, we propose a model based on rich spherical impostors, able to combine precomputed as well as dynamic procedural data, and offering seamless transitions from close instanced meshes to distant points. Our approach is architectured around an on-the-fly discrimination mechanism and intensively exploits the rough spherical geometry of the impostor proxy. In particular, we propose a new sampling mechanism to reconstruct novel views from the precomputed ones, together with a new conservative occlusion culling method, coupled with a two-pass rendering pipeline leveraging early-Z rejection. As a result, our system scales well and is even able to render sand, while supporting completely dynamic stackings.     

Online Avatar Motion Adaptation to Morphologically-similar Spaces

In avatar-mediated telepresence systems, a similar environment is assumed for involved spaces, so that the avatar in a remote space can imitate the user's motion with proper semantic intention performed in a local space. For example, touching on the desk by the user should be reproduced by the avatar in the remote space to correctly convey the intended meaning. It is unlikely, however, that the two involved physical spaces are exactly the same in terms of the size of the room or the locations of the placed objects. Therefore, a naive mapping of the user's joint motion to the avatar will not create the semantically correct motion of the avatar in relation to the remote environment. Existing studies have addressed the problem of retargeting human motions to an avatar for telepresence applications. Few studies, however, have focused on retargeting continuous full-body motions such as locomotion and object interaction motions in a unified manner. In this paper, we propose a novel motion adaptation method that allows to generate the full-body motions of a human-like avatar on-the-fly in the remote space. The proposed method handles locomotion and object interaction motions as well as smooth transitions between them according to given user actions under the condition of a bijective environment mapping between morphologically-similar spaces. Our experiments show the effectiveness of the proposed method in generating plausible and semantically correct full-body motions of an avatar in room-scale space.     

Preserving the autocovariance of texture tilings using importance sampling

By-example aperiodic tilings are popular texture synthesis techniques that allow a fast, on-the-fly generation of unbounded and non-periodic textures with an appearance matching an arbitrary input sample called the “exemplar”. But by relying on uniform random sampling, these algorithms fail to preserve the autocovariance function, resulting in correlations that do not match the ones in the exemplar. The output can then be perceived as excessively random. In this work, we present a new method which can well preserve the autocovariance function of the exemplar. It consists in fetching contents with an importance sampler taking the explicit autocovariance function as the probability density function (pdf) of the sampler. Our method can be controlled for increasing or decreasing the randomness aspect of the texture. Besides significantly improving synthesis quality for classes of textures characterized by pronounced autocovariance functions, we moreover propose a real-time tiling and blending scheme that permits the generation of high-quality textures faster than former algorithms with minimal downsides by reducing the number of texture fetches.