SCAPE: Shape Completion and Animation of People

How to read this

Category

Method / model: a data-driven statistical human body model

Contributions

• A data-driven model spanning variation in both subject shape and pose
• A pose deformation model giving non-rigid surface deformation as a function of articulated skeleton pose, plus a separate PCA shape model over 3D scans of many people
• Combines the two to synthesize realistic posed meshes for new people, enabling shape completion from sparse markers and animation from a single static scan plus marker motion

Context

A foundational parameterized body model that factors shape from pose deformation, widely built upon in later vision and animation body-model work.

Correctness

Learned from a corpus of 3D scans, so fidelity depends on how well that corpus spans target body types and poses; the pose-dependent deformation is a learned approximation rather than physical simulation, and extrapolation beyond the training distribution is a known caution.

Clarity

Conceptually clear with a clean shape/pose separation; a first pass conveys the model, a second pass clarifies the deformation learning and completion procedure.

How to read it

Worth a careful read as a foundational reference; first pass for the shape-versus-pose factorization, second pass on the deformation model and completion if you use or extend statistical body models.