Research map / 3D vision / Face and shape
AU-Blendshape for Fine-grained Stylized 3D Facial Expression Manipulation
Hao Li, Ju Dai, Feng Zhou, Kaida Ning, Lei Li, Junjun Pan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train on AU-Blendshape data (AUBlendSet) instead of D3DFACS direct AU annotations (ablation) | AUBlendNet · AUBlendSet test set | MSE_S single-AU (x1e-9): 226.41 -> 4.55 | Improved |
| train on AUBlendSet instead of CK+ (ablation) | AUBlendNet · AUBlendSet test set | MSE_M multi-AU (x1e-9): 2028.14 -> 7.62 | Improved |
| predict blendshape bases once then blend, instead of predicting each posture from AU labels (ablation) | AUBlendNet · AUBlendSet test set | inference time (s): 2.7 -> 0.3 | Improved |
| use AUBlendNet's non-autoregressive AdaLN-Zero design rather than Diffusion, VQ-VAE or Transformer generators (ablation) | AUBlendNet · AUBlendSet | MSE_S single-AU (x1e-9): 5.66 -> 4.55 (vs Diffusion) | Improved |
| add AUBlendNet expression manipulation to speech-driven models | Faceformer · MEAD | Diversity (x1e-6): 2.16 -> 2.74 | Improved |
| add AUBlendNet expression manipulation to speech-driven models | Faceformer · MEAD | LVE (x1e-4): 2.83 -> 3.04 | Hurt |
| add AUBlendNet expression manipulation to FaceDiffuser | FaceDiffuser · MEAD | LVE (x1e-4): 1.47 -> 1.76 | Hurt |
| add AUBlendNet expression manipulation to speech-driven models | Faceformer · MEAD | V-LVE (x1e-5): 2.64 -> 2.64 | No clear effect |
| use AUBlendNet-generated data augmentation for AU detection (HMP-PS) | HMP-PS · D3DFACS | F1 score: 55.04 -> 57.96 | Improved |
| use AUBlendNet-generated data augmentation for AU detection (SRERL) | SRERL · DISFA | F1 score: 55.75 -> 56.32 | Improved |
Closest papers by meaning
- AU-Aware 3D Face Reconstruction through Personalized AU-Specific Blendshape Learning (ECCV 2022 · similarity 0.90)
- AUEditNet: Dual-Branch Facial Action Unit Intensity Manipulation with Implicit Disentanglement (CVPR 2024 · similarity 0.90)
- DEEPTalk: Dynamic Emotion Embedding for Probabilistic Speech-Driven 3D Face Animation (AAAI 2025 · similarity 0.87)
- Mimic: Speaking Style Disentanglement for Speech-Driven 3D Facial Animation (AAAI 2024 · similarity 0.87)
- Learning Adaptive Spatial Coherent Correlations for Speech-Preserving Facial Expression Manipulation (CVPR 2024 · similarity 0.86)
- FG-EmoTalk: Talking Head Video Generation with Fine-Grained Controllable Facial Expressions (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
