Masquerade: Fine-Scale Details for Head-Mounted Camera Motion Capture Data

How to read this

Category

Method / production tool: fine-scale enhancement of HMC facial capture

Contributions

• Masquerade, a modular pipeline that adds fine-scale wrinkle and expression detail to sparse head-mounted-camera marker data
• A representation that separates large-scale pose (deformation gradients, so training data reuses across marker sets) from fine-scale detail (local vertex offsets)
• A region-based scheme fitting one biharmonic-kernel RBF per marker region rather than per vertex, reducing memory and compute while improving quality

Context

Addresses the gap between high-resolution seated capture systems such as Beeler et al.'s Medusa and the lower-resolution data from head-mounted cameras, combining prior large-scale/fine-scale separation approaches.Builds on: High-Quality Passive Facial Performance Capture Using Anchor Frames

Correctness

Reported as in production use at Digital Domain; this is a production-validated engineering solution rather than a benchmarked study, so quality claims rest on artist-facing use, and reconstruction still depends on the training data and marker layout.

Clarity

Clear and pragmatic; a first pass conveys the two-scale, per-region design, a second pass covers the RBF formulation.

How to read it

Focus on the large-scale vs fine-scale split and the per-region RBF choice; a second pass on the deformation-gradient representation pays off if you work with HMC capture pipelines.