Research map / Label-efficient and robust learning / Recognition and long-tail
Facial Action Unit Detection With Transformers
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 the proposed attention, multi-task embeddings, center contrastive loss, and transformer correlation model | Facial action unit detection · BP4D | F1-score (Avg.): 62.1 -> 64.2 | Helped |
| Use the proposed attention, multi-task embeddings, center contrastive loss, and transformer correlation model | Facial action unit detection · DISFA | F1-score (Avg.): 58.7 -> 61.5 | Helped |
| Use BP4D-pretrained rather than ImageNet-pretrained features (ablation) | Ours · EmotioNet validation | F1-score: 45.7 -> 47.3 | Helped |
| Pretrain on BP4D and DISFA rather than BP4D alone (ablation) | Ours · EmotioNet validation | F1-score: 47.3 -> 47.1 | Hurt |
| Add ImageNet-pretrained features to the baseline CNN (ablation) | BP4D AU detector · BP4D | F1-score: 59.0 -> 60.4 | Helped |
| Add per-AU multi-task branches (ablation) | Pretrained BP4D AU detector · BP4D | F1-score: 60.4 -> 61.0 | Helped |
| Add an attention branch (ablation) | Pretrained multi-task BP4D detector · BP4D | F1-score: 61.0 -> 62.3 | Helped |
| Add supervision from landmark-derived attention maps (ablation) | Pretrained multi-task BP4D detector with attention · BP4D | F1-score: 62.3 -> 62.5 | Helped |
| Add center contrastive loss (ablation) | Pretrained multi-task BP4D detector with supervised attention · BP4D | F1-score: 62.5 -> 62.8 | Helped |
| Add the transformer encoder (ablation) | Pretrained multi-task BP4D detector · BP4D | F1-score: 61.0 -> 63.4 | Helped |
| Use InceptionV3 rather than ResNet50 features (ablation) | Full BP4D detector · BP4D | F1-score: 63.7 -> 64.2 | Helped |
| Use WTL and WCE in both attention and inference branches rather than WTL alone (ablation) | BP4D AU detector · BP4D | F1-score: 61.3 -> 64.2 | Helped |
Closest papers by meaning
- Knowledge Augmented Deep Neural Networks for Joint Facial Expression and Action Unit Recognition (NeurIPS 2020 · similarity 0.90)
- Exploiting Semantic Embedding and Visual Feature for Facial Action Unit Detection (CVPR 2021 · similarity 0.90)
- Knowledge-Driven Self-Supervised Representation Learning for Facial Action Unit Recognition (CVPR 2022 · similarity 0.90)
- Local Relationship Learning With Person-Specific Shape Regularization for Facial Action Unit Detection (CVPR 2019 · similarity 0.89)
- Deep Region and Multi-Label Learning for Facial Action Unit Detection (CVPR 2016 · similarity 0.89)
- Action Unit Detection With Region Adaptation, Multi-Labeling Learning and Optimal Temporal Fusing (CVPR 2017 · similarity 0.88)
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.