Research map / Detection and segmentation / CNN features
Deep Adaptive Attention for Joint Facial Action Unit Detection and Face Alignment
Zhiwen Shao, Zhilei Liu, Jianfei Cai, 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 |
|---|---|---|---|
| replace region layer with hierarchical multi-scale region layer (ablation) | R-Net · BP4D AU detection | average F1-frame: 54.9 -> 55.8 | Improved |
| add Dice coefficient loss to softmax loss (ablation) | HMR-Net · BP4D AU detection | average F1-frame: 55.8 -> 56.6 | Improved |
| add per-AU loss weighting (ablation) | HMR-Net+D · BP4D AU detection | average F1-frame: 56.6 -> 57.4 | Improved |
| add face alignment module (multi-task) (ablation) | HMR-Net+DW · BP4D AU detection | average F1-frame: 57.4 -> 58.0 | Improved |
| add adaptive attention learning with alignment-based initialization (ablation) | HMR-Net+DWA · BP4D AU detection | average F1-frame: 58.0 -> 60.0 | Improved |
| add AU attention refinement (ablation) | JAA-Net · BP4D AU detection | average F1-frame: 57.4 -> 60.0 | Improved |
| add back-propagation enhancement (ablation) | JAA-Net · BP4D AU detection | average F1-frame: 59.1 -> 60.0 | Improved |
| use approximate Gaussian initial attention instead of uniform (ablation) | JAA-Net · BP4D AU detection | average F1-frame: 57.3 -> 60.0 | Improved |
| add AU detection task jointly to face alignment (ablation) | JAA-Net · BP4D face alignment | mean error: 12.23 -> 6.38 | Improved |
| add AU detection task jointly to face alignment (ablation) | JAA-Net · BP4D face alignment | failure rate: 66.85 -> 3.27 | Improved |
| add AU attention refinement (ablation) | JAA-Net · BP4D face alignment | mean error: 12.32 -> 6.38 | Improved |
| use JAA-Net instead of EAC-Net | AU detection · BP4D | average F1-frame: 55.9 -> 60.0 | Improved |
Closest papers by meaning
- JÂA-Net: Joint Facial Action Unit Detection and Face Alignment Via Adaptive Attention (IJCV 2020 · similarity 0.97)
- Action Unit Detection With Region Adaptation, Multi-Labeling Learning and Optimal Temporal Fusing (CVPR 2017 · similarity 0.90)
- EAC-Net: Deep Nets with Enhancing and Cropping for Facial Action Unit Detection (TPAMI 2018 · similarity 0.89)
- Exploiting Semantic Embedding and Visual Feature for Facial Action Unit Detection (CVPR 2021 · similarity 0.88)
- Deep Region and Multi-Label Learning for Facial Action Unit Detection (CVPR 2016 · similarity 0.88)
- Trend-Aware Supervision: On Learning Invariance for Semi-supervised Facial Action Unit Intensity Estimation (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.
