← ArchivePaper2015
SMPL: A Skinned Multi-Person Linear Model
Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, Michael J. Black
SMPL body model factoring shape and pose-dependent deformations for efficient synthesis of realistic human body shapes and animations.
Abstract
We present a learned model of human body shape and pose-dependent shape variation that is more accurate than previous models and is compatible with existing graphics pipelines. Our Skinned Multi-Person Linear model (SMPL) is a skinned vertex-based model that accurately represents a wide variety of body shapes in natural human poses. The parameters of the model are learned from data including the rest pose template, blend weights, pose-dependent blend shapes, identity-dependent blend shapes, and a regressor from vertices to joint locations. Unlike previous models, the pose-dependent blend shapes are a linear function of the elements of the pose rotation matrices. This simple formulation enables training the entire model from a relatively large number of aligned 3D meshes of different people in different poses. We quantitatively evaluate variants of SMPL using linear or dual-quaternion blend skinning and show that both are more accurate than a Blend-SCAPE model trained on the same data. We also extend SMPL to realistically model dynamic soft-tissue deformations. Because it is based on blend skinning, SMPL is compatible with existing rendering engines and we make it available for research purposes.
How to read this
- Category
- Method / model: learned parametric body model (SMPL)
- Contributions
- Presents a skinned vertex-based human body model that factors identity-dependent and pose-dependent shape variation and is compatible with existing graphics pipelines.
- Makes pose-dependent blend shapes a linear function of the elements of the pose rotation matrices, enabling training the whole model (template, blend weights, blend shapes, joint regressor) from many aligned 3D meshes.
- Shows SMPL with linear or dual-quaternion blend skinning is more accurate than a Blend-SCAPE model on the same data, and extends to dynamic soft-tissue deformation.
- Context
- A blend-skinning successor to deformation-based body models such as Anguelov et al.'s SCAPE, designed to stay compatible with standard rendering engines.Builds on: SCAPE: Shape Completion and Animation of People
- Correctness
- Accuracy claims are relative to a Blend-SCAPE baseline trained on the same data and depend on the quality and coverage of the aligned mesh training set; the linear pose-blendshape formulation is a deliberate simplification that trades some expressiveness for trainability and pipeline compatibility.
- Clarity
- Accessible given a skinning background; a first pass conveys the factored model, a careful second pass for the blend-shape parameterization and training.
- How to read it
- First pass for the shape/pose factorization and why linear pose blendshapes matter; second pass on the formulation and training if using or extending SMPL.
Builds on
Built upon by
- Dyna: A Model of Dynamic Human Shape in Motion 2015
- ClothCap: Seamless 4D Clothing Capture and Retargeting 2017
- AMASS: Archive of Motion Capture as Surface Shapes 2019
- Expressive Body Capture: 3D Hands, Face, and Body from a Single Image 2019
- Learning-Based Animation of Clothing for Virtual Try-On 2019
- Learning to Dress 3D People in Generative Clothing 2020
- NiLBS: Neural Inverse Linear Blend Skinning 2020
- SoftSMPL: Data-driven Modeling of Nonlinear Soft-tissue Dynamics for Parametric Humans 2020
- STAR: Sparse Trained Articulated Human Body Regressor 2020
- Animatable Neural Radiance Fields for Modeling Dynamic Human Bodies 2021
- Learning Skeletal Articulations with Neural Blend Shapes 2021
- Neural Body: Implicit Neural Representations with Structured Latent Codes for Novel View Synthesis of Dynamic Humans 2021
- S3: Neural Shape, Skeleton, and Skinning Fields for 3D Human Modeling 2021
- SCANimate: Weakly Supervised Learning of Skinned Clothed Avatar Networks 2021
- SNARF: Differentiable Forward Skinning for Animating Non-Rigid Neural Implicit Shapes 2021
- BOSS: Bones, Organs and Skin Shape Model 2023
- From Skin to Skeleton: Towards Biomechanically Accurate 3D Digital Humans 2023
- HIT: Estimating Internal Human Implicit Tissues from the Body Surface 2024
- TailorMe: Self-Supervised Learning of an Anatomically Constrained Volumetric Human Shape Model 2024
- ATLAS: Decoupling Skeletal and Shape Parameters for Expressive Parametric Human Modeling 2025
- MHR: Momentum Human Rig 2025
- SOMA: Unifying Parametric Human Body Models 2026
Related work
- ClothCap: Seamless 4D Clothing Capture and Retargeting 2017 / SIGGRAPH
- Expressive Body Capture: 3D Hands, Face, and Body from a Single Image 2019 / CVPR
- STAR: Sparse Trained Articulated Human Body Regressor 2020 / Eurographics
- 3D Mesh Pose Transfer Based on Skeletal Deformation 2023 / CASA
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 →