← ArchivePaper2021

HeterSkinNet: A Heterogeneous Network for Skin Weights Prediction

Xiaoyu Pan, Jiancong Huang, Jiaming Mai, He Wang, Honglin Li, Tongkui Su, Wenjun Wang, Xiaogang Jin

I3DAcademic8 citesSkinning ML Deformation

Builds a heterogeneous graph over mesh vertices and skeletal bones with a HollowDist metric, predicting production-quality skin weights for arbitrary character topologies.

Open the paper ↗Trace in the lineage graph

Abstract

Character rigging is universally needed in computer graphics but notoriously laborious. We present a new method, HeterSkinNet, aiming to fully automate such processes and significantly boost productivity. Given a character mesh and skeleton as input, our method builds a heterogeneous graph that treats the mesh vertices and the skeletal bones as nodes of different types and uses graph convolutions to learn their relationships. To tackle the graph heterogeneity, we propose a new graph network convolution operator that transfers information between heterogeneous nodes. The convolution is based on a new distance HollowDist that quantifies the relations between mesh vertices and bones. We show that HeterSkinNet is robust for production characters by providing the ability to incorporate meshes and skeletons with arbitrary topologies and morphologies (e.g., out-of-body bones, disconnected mesh components, etc.). Through exhaustive comparisons, we show that HeterSkinNet outperforms state-of-the-art methods by large margins in terms of rigging accuracy and naturalness. HeterSkinNet provides a solution for effective and robust character rigging.

How to read this

Category: Method: learning-based automatic skin weight prediction
Contributions: HeterSkinNet, a heterogeneous graph over mesh vertices and skeletal bones as distinct node types, using graph convolutions to learn their relationships
A new graph convolution operator that transfers information between heterogeneous nodes
A new HollowDist distance quantifying vertex-to-bone relations, supporting arbitrary mesh and skeleton topologies including out-of-body bones and disconnected components
Context: Continues deep-learning automatic skinning for production characters, building on graph-network skin binding such as Liu et al.'s NeuroSkinning.Builds on: NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks
Correctness: Reported to be robust on production characters and to outperform prior methods on rigging accuracy and naturalness through comparisons; as with learned skinning, generalization depends on the training characters and the chosen distance metric.
Clarity: Accessible (I3D); a first pass conveys the heterogeneous-graph idea, a second pass clarifies the HollowDist metric and convolution operator.
How to read it: Read for the heterogeneous-graph formulation and HollowDist; second pass on the convolution operator if you work on automatic rigging.

Builds on

NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks 2019

Built upon by

Nothing yet.

Related work

NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks 2019 / SIGGRAPH
Learning Skeletal Articulations with Neural Blend Shapes 2021 / SIGGRAPH
Real-Time Skeletal Skinning with Optimized Centers of Rotation 2016 / SIGGRAPH
A Statistical Model of Human Pose and Body Shape 2009 / CGF

Keywords

heterskinnet skin weights prediction heterogeneous graph neural network graph convolution hollowdist character rigging linear blend skinning dual quaternion skinning mesh graph skeleton graph edgeconv surface voxelization breadth-first search out-of-body bones vertex-bone distance neuroskinning rignet kullback-leibler divergence

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 →