SMPL

SMPL, short for Skinned Multi-Person Linear model, is a parametric 3D human body model used to represent body shape and pose with a compact set of parameters. Introduced by researchers from the Max Planck Institute for Informatics in 2015, it provides a consistent, differentiable mesh of the unclothed human body that can be driven by low-dimensional inputs.

The model uses a fixed topology: a triangular mesh with 6890 vertices and 23 joints. Shape is encoded by a 10-dimensional vector β, while pose is encoded by a 72-dimensional vector θ (typically representing three-axis rotations for multiple joints, plus global orientation). A separate 3D translation accounts for the overall position. The output is a mesh M(β, θ) obtained through a linear blend skinning framework, augmented by pose-dependent corrective terms to capture soft-tissue deformations that occur during articulation.

SMPL is designed to be fitted to imagery or point-cloud data, enabling recovery of detailed 3D human shape and pose from single or multi-view observations. It has become a standard building block in computer vision and graphics pipelines for tasks such as 3D human pose estimation, animation, and virtual reality applications. The model is extendable; variants include SMPL-X, which adds detailed hands and facial geometry, and SMPL+H, which emphasizes hand details. There are also tools and methods for fitting SMPL to images or scans, and for integrating SMPL with clothing or other accessories.

Limitations include its fixed topology, which makes accurate clothing modeling challenging, and potential biases from training data. While powerful for unclothed body representation, SMPL is not a clothing-aware model by default and may require extensions or alternative methods for fully clothed scenarios.