Research map / Label-efficient and robust learning / Self-supervised and contrastive
Knowledge-Driven Self-Supervised Representation Learning for Facial Action Unit Recognition
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 FACS-guided local representations and contrastive and predicting learning instead of generic self-supervised features | AU recognition · BP4D | F1 score (Avg.): 56.1 -> 61.6 | Helped |
| Use FACS-guided local representations and contrastive and predicting learning instead of generic self-supervised features | AU recognition · DISFA | F1 score (Avg.): 45.0 -> 60.2 | Helped |
| Use the proposed learned representations and fine-tune the encoder | AU recognition · BP4D | F1 score (Avg.): 64.1 -> 64.5 | Helped |
| Use the proposed learned representations and fine-tune the encoder | AU recognition · DISFA | F1 score (Avg.): 63.8 -> 64.5 | Helped |
| Add intra-area instance discrimination loss (λ = 0.1) to the inter-area contrastive objective (ablation) | Knowledge-driven self-supervised AU representation learning · BP4D | F1 score: increases by 7.6% | Helped |
| Add intra-area instance discrimination loss (λ = 0.1) to the inter-area contrastive objective (ablation) | Knowledge-driven self-supervised AU representation learning · DISFA | F1 score: increases by 8.6% | Helped |
| Increase the intra-area loss weight beyond λ = 0.1 (ablation) | Knowledge-driven self-supervised AU representation learning · BP4D and DISFA | F1 score | Hurt |
| Add the contrastive learning component to predicting learning (α = 0.01) (ablation) | Knowledge-driven self-supervised AU representation learning · BP4D and DISFA | F1 score | Helped |
| Increase the contrastive component weight beyond α = 0.01 (ablation) | Knowledge-driven self-supervised AU representation learning · BP4D and DISFA | F1 score | Hurt |
Closest papers by meaning
- Semantic Relationships Guided Representation Learning for Facial Action Unit Recognition (AAAI 2019 · similarity 0.92)
- Weakly Supervised Facial Action Unit Recognition Through Adversarial Training (CVPR 2018 · similarity 0.90)
- Facial Action Unit Detection With Transformers (CVPR 2021 · similarity 0.90)
- Exploiting Semantic Embedding and Visual Feature for Facial Action Unit Detection (CVPR 2021 · similarity 0.89)
- Self-Supervised Facial Representation Learning with Facial Region Awareness (CVPR 2024 · similarity 0.87)
- PrefAce: Face-Centric Pretraining with Self-Structure Aware Distillation (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.