Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-Label Compound Expression Recognition: C-EXPR Database & Network
Read the paperCheck 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 |
|---|---|---|---|
| Pretrain C-EXPR-NET on C-EXPR-DB before RAF-DB fine-tuning | C-EXPR-NET · RAF-DB compound expressions | UAR: 0.553 -> 0.601 | Helped |
| Use compound-expression training for zero-shot basic expression recognition | C-EXPR-NET · RAF-DB basic expressions | UAR | Helped |
| Formulate CER as multi-label softmax with KL-divergence rather than multi-class (ablation) | ResNet expression branch · C-EXPR-DB compound expressions | F1: 0.39 -> 0.46 | Helped |
| Use softmax with KL-divergence rather than sigmoid cross-entropy for multi-label CER (ablation) | ResNet expression branch · C-EXPR-DB compound expressions | F1: 0.40 -> 0.46 | Helped |
| Add basic-expression annotated images as data augmentation (ablation) | ResNet expression branch · RAF-DB compound expressions | UAR: 0.43 -> 0.47 | Helped |
| Add AU detection as an auxiliary task (ablation) | ResNet expression branch · RAF-DB compound expressions | UAR: 0.47 -> 0.52 | Helped |
| Add AU detection as an auxiliary task (ablation) | ResNet expression branch · C-EXPR-DB compound expressions | F1: 0.48 -> 0.55 | Helped |
| Add distribution matching between expression-derived and predicted AU distributions (ablation) | C-EXPR-NET · RAF-DB compound expressions | UAR: 0.52 -> 0.55 | Helped |
| Add distribution matching between expression-derived and predicted AU distributions (ablation) | C-EXPR-NET · C-EXPR-DB compound expressions | F1: 0.55 -> 0.60 | Helped |
| Add distribution matching to multi-task AU detection (ablation) | C-EXPR-NET · RAF-DB AU detection | F1: 0.56 -> 0.60 | Helped |
| Add distribution matching to multi-task AU detection (ablation) | C-EXPR-NET · C-EXPR-DB AU detection | F1: 0.51 -> 0.55 | Helped |
| Add AU semantic descriptions, semantic encoding, and cross-modality attention (ablation) | ResNet50 AU detector · C-EXPR-DB AU detection | F1: 0.49 -> 0.52 | Helped |
Closest papers by meaning
- A Unified and Interpretable Emotion Representation and Expression Generation (CVPR 2024 · similarity 0.86)
- Navigating Label Ambiguity for Facial Expression Recognition in the Wild (AAAI 2025 · similarity 0.86)
- Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis (AAAI 2026 · similarity 0.85)
- CARAT: Contrastive Feature Reconstruction and Aggregation for Multi-Modal Multi-Label Emotion Recognition (AAAI 2024 · similarity 0.85)
- Open-Set Facial Expression Recognition (AAAI 2024 · similarity 0.85)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · 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.