Skip to content

← ArchivePaper2025

HumanRig: Learning Automatic Rigging for Humanoid Character in a Large Scale Dataset

Zedong Chu, Feng Xiong, Meiduo Liu, Jinzhi Zhang, Mingqi Shao, Zhaoxu Sun, Di Wang, Mu Xu

CVPRAcademic16 citesRiggingML Deformation

Introduces the first large-scale humanoid rigging dataset of 11,434 T-posed meshes and a transformer-based framework fusing skeleton priors with 3D mesh features for automatic rigging.

Abstract

With the rapid evolution of 3D generation algorithms, the cost of producing 3D humanoid character models has plummeted, yet the field is impeded by the lack of a comprehensive dataset for automatic rigging, a pivotal step in character animation. Addressing this gap, we present HumanRig, the first large-scale dataset specifically designed for 3D humanoid character rigging, encompassing 11,434 meticulously curated T-posed meshes adhered to a uniform skeleton topology. Capitalizing on this dataset, we introduce an innovative, data-driven automatic rigging framework, which overcomes the limitations of GNNbased methods in handling complex AI-generated meshes. Our approach integrates a Prior-Guided Skeleton Estimator (PGSE) module, which uses 2D skeleton joints to provide a preliminary 3D skeleton, and a Mesh-Skeleton Mutual Attention Network (MSMAN) that fuses skeleton features with 3D mesh features extracted by a U-shaped point transformer. This enables a coarse-to-fine 3D skeleton joint regression and a robust skinning estimation, surpassing previous methods in quality and versatility. This work not only remedies the dataset deficiency in rigging research but also propels the animation industry towards more efficient and automated character rigging pipelines.

How to read this

Category
Dataset + method: automatic humanoid rigging
Contributions
  • HumanRig, presented as the first large-scale humanoid rigging dataset of 11,434 curated T-posed meshes sharing a uniform skeleton topology.
  • A data-driven rigging framework with a Prior-Guided Skeleton Estimator (PGSE) that uses 2D skeleton joints to seed a 3D skeleton.
  • A Mesh-Skeleton Mutual Attention Network (MSMAN) fusing skeleton and U-shaped point-transformer mesh features for coarse-to-fine joint regression and skinning.
Context
Extends neural auto-rigging beyond GNN-based methods such as RigNet, targeting the complex AI-generated meshes that earlier approaches struggled with.Builds on: RigNet: Neural Rigging for Articulated Characters
Correctness
Reported to surpass prior methods in quality and versatility, but the uniform skeleton topology and T-pose assumption define the scope, so a reader should keep in mind generalization to non-standard skeletons, non-T-poses, or non-humanoid shapes is outside the dataset's design.
Clarity
Reasonably accessible architecture story; a first pass conveys the two-module pipeline, a second pass is needed for the PGSE/MSMAN details.
How to read it
Skim the dataset construction and the PGSE-then-MSMAN pipeline first; second pass the attention fusion and skinning estimation if implementing or comparing against RigNet.

Builds on

Built upon by

Nothing yet.

Related work

Keywords

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 →