A large-scale collection of 3D humanoid character meshes and their corresponding skeletal rigging data. The dataset provides diverse character models paired with joint hierarchies and skinning weights to facilitate research in automatic rigging for humanoid characters.
Use Cases
- Train neural networks to predict skeletal joint positions from 3D mesh geometry.
- Develop automatic skinning weight estimation models using vertex-to-joint mapping data.
- Benchmark the performance of rigging algorithms across a wide variety of humanoid character designs.
Strengths
- Includes 3D humanoid character meshes with diverse geometric topologies.
- Features ground-truth skeletal joint hierarchies for each character model.
- Provides vertex-wise skinning weights for character deformation modeling.
- Associated with the CVPR2025 paper 'HumanRig: Learning Automatic Rigging for Humanoid Character in a Large Scale Dataset'.