Research map / Probabilistic and scientific modeling / Causal discovery
Facial Action Unit Detection by Adaptively Constraining Self-Attention and Causally Deconfounding Sample
Zhiwen Shao, Hancheng Zhu, Yong Zheng Zhou, Xiang Xiang, Bing Liu, Rui Yao, Lizhuang Ma
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 |
|---|---|---|---|
| add adaptive constraining of average self-attention (ablation) | B-Net · BP4D | Avg F1-frame: 60.2 -> 62.2 | Improved |
| constrain each self-attention channel instead of the average (ablation) | A^v-Net · BP4D | Avg F1-frame: 62.2 -> 61.9 | Hurt |
| add causal deconfounding with per-batch prototypes in each AU branch (ablation) | A^v-Net · BP4D | Avg F1-frame: 62.2 -> 64.6 | Improved |
| compute sample prototypes per mini-batch rather than per epoch (ablation) | AC2D · BP4D | Avg F1-frame: 63.4 -> 64.6 | Improved |
| apply deconfounding per AU branch rather than on shared feature (ablation) | AC2D · BP4D | Avg F1-frame: 60.0 -> 64.6 | Improved |
| add sample deconfounding (BP4D to GFT) (ablation) | A^v-Net · BP4D to GFT cross-dataset | Avg F1-frame: 37.0 -> 42.3 | Improved |
| add sample deconfounding (GFT to BP4D) (ablation) | A^v-Net · GFT to BP4D cross-dataset | Avg F1-frame: 33.9 -> 38.4 | Improved |
Closest papers by meaning
- Breaking Spurious Correlations: Uncertainty-Driven Causal Transformers for AU Detection (CVPR 2026 · similarity 0.91)
- JÂA-Net: Joint Facial Action Unit Detection and Face Alignment Via Adaptive Attention (IJCV 2020 · similarity 0.89)
- Causal Intervention for Subject-Deconfounded Facial Action Unit Recognition (AAAI 2022 · similarity 0.89)
- Context-Aware Feature and Label Fusion for Facial Action Unit Intensity Estimation With Partially Labeled Data (ICCV 2019 · similarity 0.88)
- Trend-Aware Supervision: On Learning Invariance for Semi-supervised Facial Action Unit Intensity Estimation (AAAI 2024 · similarity 0.86)
- AUEditNet: Dual-Branch Facial Action Unit Intensity Manipulation with Implicit Disentanglement (CVPR 2024 · similarity 0.86)
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.
