Reconstructing Personalized Anatomical Models for Physics-based Body Animation

How to read this

Category

Method: personalized anatomical model reconstruction for physics-based body animation

Contributions

• Builds personalized physics-ready anatomical models from only a set of 3D surface scans, starting from an average-male template
• A large-scale optimization that jointly solves for subject-specific parameters (bone, muscle, adipose shape) and pose-dependent parameters so the model explains the captured scans
• Produces realistic internal bone and muscle geometry that naturally supports inertia, gravity, and collisions under Newtonian dynamics

Context

Builds on anatomy-template fitting in the lineage of Dicko et al.'s Anatomy Transfer, going beyond surface-only data-driven body modeling toward physics-based internal structure.Builds on: Anatomy Transfer

Correctness

Validated by how closely the reconstructed model explains the captured multi-pose scans; readers should keep in mind it starts from an average-male template and infers internal anatomy from surface evidence, so internal structures are plausible estimates rather than measured ground truth.

Clarity

Technical; a first pass conveys the scan-to-anatomy goal and the joint-optimization framing, while the optimization details need a careful second pass.

How to read it

First pass for the surface-scan-to-physics-model pipeline and the joint subject/pose optimization idea; second pass on the optimization formulation if you need to reproduce or extend the fitting.