← ArchivePaper2021

Learning Skeletal Articulations with Neural Blend Shapes

Peizhuo Li, Kfir Aberman, Rana Hanocka, Libin Liu, Olga Sorkine-Hornung, Baoquan Chen

Neural blend shapes learned jointly with skeleton articulations, producing pose-dependent non-linear deformations from a single neutral-pose mesh.

Open the paper ↗Trace in the lineage graph

Abstract

Animating a newly designed character using motion capture (mocap) data is a long standing problem in computer animation. A key consideration is the skeletal structure that should correspond to the available mocap data, and the shape deformation in the joint regions, which often requires a tailored, pose-specific refinement. In this work, we develop a neural technique for articulating 3D characters using enveloping with a pre-defined skeletal structure which produces high quality pose dependent deformations. Our framework learns to rig and skin characters with the same articulation structure ( e.g. , bipeds or quadrupeds), and builds the desired skeleton hierarchy into the network architecture. Furthermore , we propose neural blend shapes - a set of corrective pose-dependent shapes which improve the deformation quality in the joint regions in order to address the notorious artifacts resulting from standard rigging and skinning. Our system estimates neural blend shapes for input meshes with arbitrary connectivity, as well as weighting coefficients which are conditioned on the input joint rotations. Unlike recent deep learning techniques which supervise the network with ground-truth rigging and skinning parameters, our approach does not assume that the training data has a specific underlying deformation model.

How to read this

Category: Method: learned rigging and skinning with neural blend shapes
Contributions: A neural technique that rigs and skins characters of a shared articulation structure (bipeds, quadrupeds) with a predefined skeleton baked into the network architecture
Neural blend shapes, a set of corrective pose-dependent shapes that improve joint-region deformation and reduce standard skinning artifacts
Estimation of blend shapes and weighting coefficients conditioned on joint rotations for meshes with arbitrary connectivity
Context: Sits in the learned-rigging and pose-dependent-deformation lineage and relates to parametric body modelling such as SMPL (Loper et al.), which similarly uses pose-corrective shapes on a skinned model.Builds on: SMPL: A Skinned Multi-Person Linear Model
Correctness: Assumes a fixed articulation structure and a single neutral-pose input; the method targets joint-region artifact reduction, so readers should note its scope is characters matching the trained skeleton class rather than arbitrary topologies or skeletons.
Clarity: Accessible motivation (rigging/skinning artifacts); a first pass conveys the neural-blend-shape idea, a second pass clarifies the architecture and conditioning.
How to read it: Read pass one for the rig-and-skin-plus-corrective concept; do a second pass on the network architecture and how blend shapes are conditioned on rotations if you want to implement or compare against it.

Builds on

SMPL: A Skinned Multi-Person Linear Model 2015

Built upon by

Nothing yet.

Related work

NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks 2019 / SIGGRAPH
Real-Time Deformation with Coupled Cages and Skeletons 2020 / SCA
HeterSkinNet: A Heterogeneous Network for Skin Weights Prediction 2021 / I3D
S3: Neural Shape, Skeleton, and Skinning Fields for 3D Human Modeling 2021 / CVPR

Keywords

neural blend shapes linear blend skinning dual quaternion skinning rigging skinning weights envelope deformation residual deformation pose-dependent deformation meshcnn skeleton-aware network indirect supervision smpl mocap arbitrary mesh connectivity forward kinematics rignet

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →