← ArchivePaper2020

Accurate Face Rig Approximation with Deep Differential Subspace Reconstruction

SIGGRAPHIndustrial29 cites1 descendantFacial ML Deformation

Deep network for approximating complex face rig computations using differential subspace reconstruction for real-time face deformation.

Open the paper ↗Trace in the lineage graph

Abstract

To be suitable for film-quality animation, rigs for character deformation must fulfill a broad set of requirements. They must be able to create highly stylized deformation, allow a wide variety of controls to permit artistic freedom, and accurately reflect the design intent. Facial deformation is especially challenging due to its nonlinearity with respect to the animation controls and its additional precision requirements, which often leads to highly complex face rigs that are not generalizable to other characters. This lack of generality creates a need for approximation methods that encode the deformation in simpler structures. We propose a rig approximation method that addresses these issues by learning localized shape information in differential coordinates and, separately, a subspace for mesh reconstruction. The use of differential coordinates produces a smooth distribution of errors in the resulting deformed surface, while the learned subspace provides constraints that reduce the low frequency error in the reconstruction. Our method can reconstruct both face and body deformations with high fidelity and does not require a set of well-posed animation examples, as we demonstrate with a variety of production characters.

How to read this

Category: Method: neural approximation of a film-quality face rig
Contributions: A rig approximation that learns localized shape information in differential coordinates
A separately learned subspace for mesh reconstruction that constrains and reduces low-frequency error
High-fidelity reconstruction of both face and body deformation without requiring a set of well-posed animation examples
Context: Relates to learned rig/deformation approximation (referenced Fast and Deep Deformation Approximations, Bailey 2018), targeting the nonlinearity and precision demands specific to facial rigs.Builds on: Fast and Deep Deformation Approximations
Correctness: Differential coordinates are used to smooth the error distribution and the learned subspace to curb low-frequency error; it is an approximation of a complex rig, so output fidelity is bounded by training coverage and the chosen subspace, and per-character generality of the rig itself remains a stated motivation rather than a solved problem.
Clarity: Moderately technical; a first pass conveys the differential-coordinates-plus-subspace split, a second pass is needed for the learning and reconstruction formulation.
How to read it: On the first pass focus on why differential coordinates plus a reconstruction subspace address facial nonlinearity and error distribution; second pass on the training setup and error analysis if approximating your own rigs for real-time use.

Builds on

Fast and Deep Deformation Approximations 2018

Built upon by

CANRIG: Cross-Attention Neural Face Rigging with Variable Local Control 2026

Related work

FaceBaker: Baking Character Facial Rigs with Machine Learning 2020 / SIGGRAPH
Fast and Deep Facial Deformations 2020 / SIGGRAPH
Implementing a Machine Learning Deformer for CG Crowds: Our Journey 2024 / DigiPro
A Facial Composite Editor for Blendshape Characters 2012 / DigiPro

Keywords

rig approximation facial deformation differential coordinates subspace reconstruction deep learning neural network laplacian operator anchor points linear blend skinning nonlinear deformation blendshapes mesh reconstruction pca cholesky factorization face rig body deformation delta space crowd simulation

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →