← ArchivePaper2020
Facial Expression Synthesis using a Global-Local Multilinear Framework
Mengjiao Wang, Derek Bradley, Stefanos Zafeiriou, Thabo Beeler
Global-local multilinear model synthesizes identity-preserving facial expressions that extrapolate well beyond the training data pool.
Abstract
We present a practical method to synthesize plausible 3D facial expressions for a particular target subject. The ability to synthesize an entire facial rig from a single neutral expression has a large range of applications both in computer graphics and computer vision, ranging from the efficient and cost‐effective creation of CG characters to scalable data generation for machine learning purposes. Unlike previous methods based on multilinear models, the proposed approach is capable to extrapolate well outside the sample pool, which allows it to plausibly predict the identity of the target subject and create artifact free expression shapes while requiring only a small input dataset. We introduce global‐local multilinear models that leverage the strengths of expression‐specific and identity‐specific local models combined with coarse motion estimations from a global model. Experimental results show that we achieve high‐quality, plausible facial expression synthesis results for an individual that outperform existing methods both quantitatively and qualitatively.
How to read this
- Category
- Method: facial expression synthesis from a multilinear model
- Contributions
- Synthesizes a plausible 3D facial rig for a target subject from a single neutral expression
- Introduces global-local multilinear models combining expression- and identity-specific local models with a coarse global motion estimate
- Extrapolates beyond the training sample pool while requiring only a small input dataset
- Context
- Builds directly on the multilinear face-modeling tradition, notably Vlasic et al.'s Face Transfer with Multilinear Models, while addressing that family's poor extrapolation.Builds on: Face Transfer with Multilinear Models
- Correctness
- Reported to outperform prior multilinear methods quantitatively and qualitatively, but claims rest on producing plausible, identity-preserving shapes from limited input; readers should treat the synthesized rig as a plausible prediction rather than a captured ground truth and watch how far extrapolation holds.
- Clarity
- Accessible; a first pass conveys the global-local idea, a second pass clarifies the tensor formulation.
- How to read it
- First pass for the global-local decomposition intuition; second pass on the multilinear math and the extrapolation experiments if comparing against standard multilinear baselines.
Builds on
Built upon by
Nothing yet.
Related work
Keywords
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 →