Key points
- SMPL is the parametric human body model that most learning-based pose and motion methods output.
- SMPL-X extends it with hands and face for expressive whole-body capture.
- Text-to-motion models including MDM produce SMPL parameters, so the representation is prerequisite to reading their outputs.
Summary
Loper et al. (2015) introduced SMPL, a parametric body model that represents a human figure as a blend of shape and pose parameters derived from a large database of body scans; the model outputs a posed mesh via learned blend shapes and skinning weights. Pavlakos et al. (2019) extended it to SMPL-X, adding articulated hands and an expressive face model for whole-body capture. Together they form the de facto output representation for learning-based human pose estimation, motion capture and text-to-motion generation: the Motion Diffusion Model, DeepPhase and most motion-synthesis methods produce SMPL parameters, making the model prerequisite knowledge for reading their outputs and results in animation research and teaching.
Related items
- Human Motion Diffusion Model (MDM)
- DeepPhase: Periodic Autoencoders for Learning Motion Phase Manifolds
- A deep learning framework for character motion synthesis and editing (SIGGRAPH 2016 / SIGGRAPH 2026 Test-of-Time award)
Source
Source: ACM Transactions on Graphics (SIGGRAPH Asia 2015) / CVPR 2019 ↗ (Research)
Cite this item
Loper, M., Mahmood, N., Romero, J., Pons-Moll, G. & Black, M. J. (2015). ‘SMPL and SMPL-X parametric body models’, ACM Transactions on Graphics (SIGGRAPH Asia 2015) / CVPR 2019. Available at: https://dl.acm.org/doi/10.1145/2816795.2818013
Your reference manager can also read this page directly: with the Zotero (or Mendeley) browser connector installed, save it straight to your library. Whole-collection exports: RIS, BibTeX, CSL-JSON.