Research map / Detection and segmentation / CNN features
Local Relationship Learning With Person-Specific Shape Regularization for Facial Action Unit Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use LP-Net instead of JAA-Net | AU detection · BP4D | Avg. F1-frame: 60.0 -> 61.0 | Helped |
| Use LP-Net instead of JAA-Net | AU detection · DISFA | Avg. F1-frame: 56.0 -> 56.9 | Helped |
| Use ResNet-34 instead of ResNet-18 as the stem network (ablation) | BP4D AU detection · BP4D | Avg. F1-frame: 52.9 -> 53.7 | Helped |
| Use ResNet-50 instead of ResNet-34 as the stem network (ablation) | BP4D AU detection · BP4D | Avg. F1-frame: 53.7 -> 52.5 | Hurt |
| Add Selective Learning balancing (ablation) | ResNet-34 AU detector · BP4D | Avg. F1-frame: 53.7 -> 55.2 | Helped |
| Add data augmentation (ablation) | ResNet-34+SL · BP4D | Avg. F1-frame: 55.2 -> 56.5 | Helped |
| Pre-train the model on VGGFace2 (ablation) | ResNet-34+SL/DA · BP4D | Avg. F1-frame: 56.5 -> 58.0 | Helped |
| Use local features for AU prediction and mean-pool their predictions (ablation) | Stem-Net · BP4D | Avg. F1-frame: 58.0 -> 58.8 | Helped |
| Add local relationship learning over local features (ablation) | Stem-Net · BP4D | Avg. F1-frame: 58.0 -> 60.2 | Helped |
| Use relationship learning on FACS-defined local areas rather than all local features (ablation) | Stem-Net · BP4D | Avg. F1-frame: 58.8 -> 59.0 | Mixed |
| Add P-Net shape features without regularization loss Lr (ablation) | Stem-Net · BP4D | Avg. F1-frame: 58.0 -> 58.7 | Helped |
| Add regularization loss Lr to separate AU features from face-shape features (ablation) | Stem-Net+P-Net w/o Lr · BP4D | Avg. F1-frame: 58.7 -> 59.8 | Helped |
Closest papers by meaning
- Facial Action Unit Detection With Transformers (CVPR 2021 · similarity 0.89)
- Knowledge-Driven Self-Supervised Representation Learning for Facial Action Unit Recognition (CVPR 2022 · similarity 0.89)
- Semantic Relationships Guided Representation Learning for Facial Action Unit Recognition (AAAI 2019 · similarity 0.89)
- Exploiting Semantic Embedding and Visual Feature for Facial Action Unit Detection (CVPR 2021 · similarity 0.88)
- Knowledge Augmented Deep Neural Networks for Joint Facial Expression and Action Unit Recognition (NeurIPS 2020 · similarity 0.88)
- 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.