Research map / Label-efficient and robust learning / Recognition and long-tail
Uncertainty-aware facial expression recognition
Models label ambiguity and bias for facial expression recognition.
41 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Recognition and long-tail, keyword cluster “face · person identification · facial”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 1 tracked benchmark.
12 papers in 2023–24 and 5 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 1.0% to 0.4%.
Open questions the data points to
- RAF-DB (accuracy): best went from 89.47 (up to 2024) to 93.02 in 2025 — QCS:Feature Refining from Quadruplet Cross Similarity for Facial Expression Recognition (AAAI 2025 · 1 citation)
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · EmotioNet: An Accurate, Real-Time Algorithm for the Automatic Annotation of a Million Facial Expressions in the Wild (CVPR 2016 · 632 citations)
- 2017 · Reliable Crowdsourcing and Deep Locality-Preserving Learning for Expression Recognition in the Wild (CVPR 2017 · 1,404 citations)
- 2018 · Weakly-Supervised Deep Convolutional Neural Network Learning for Facial Action Unit Intensity Estimation (CVPR 2018 · 59 citations)
- 2019 · Facial Emotion Distribution Learning by Exploiting Low-Rank Label Correlations Locally (CVPR 2019 · 96 citations)
- 2020 · Suppressing Uncertainties for Large-Scale Facial Expression Recognition (CVPR 2020 · 722 citations)
- 2021 · Robust Lightweight Facial Expression Recognition Network with Label Distribution Training (AAAI 2021 · 283 citations)
- 2022 · Face2Exp: Combating Data Biases for Facial Expression Recognition (CVPR 2022 · 138 citations)
- 2023 · Micron-BERT: BERT-Based Facial Micro-Expression Recognition (CVPR 2023 · 133 citations)
- 2024 · OpticalDR: A Deep Optical Imaging Model for Privacy-Protective Depression Recognition (CVPR 2024 · 14 citations)
- 2025 · SSAN: A Symbol Spatial-Aware Network for Handwritten Mathematical Expression Recognition (AAAI 2025 · 7 citations)
- 2026 · Behaviour4All: A Dependency-Aware Toolkit for in-the-wild Facial Behaviour Analysis (NeurIPS 2026)
Earliest papers here
- EmotioNet: An Accurate, Real-Time Algorithm for the Automatic Annotation of a Million Facial Expressions in the Wild (CVPR 2016 · 632 citations)
- Facial Expression Intensity Estimation Using Ordinal Information (CVPR 2016 · 110 citations)
- Reliable Crowdsourcing and Deep Locality-Preserving Learning for Expression Recognition in the Wild (CVPR 2017 · 1,404 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
RAF-DB · accuracy (higher is better)
5 papers, 2020–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 88.14 | Suppressing Uncertainties for Large-Scale Facial Expression Recognition (CVPR 2020 · 722 citations) |
| 2021 | 89.47 | Feature Decomposition and Reconstruction Learning for Effective Facial Expression Recognition (CVPR 2021 · 213 citations) |
| 2025 | 93.02 | QCS:Feature Refining from Quadruplet Cross Similarity for Facial Expression Recognition (AAAI 2025 · 1 citation) |
What has been tried
39 papers here report 405 measured changes (a component added, removed or swapped, compared with a baseline): helped 79%, no clear effect 2%, hurt 15%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add emotion-task pretraining before depression training | AVEC 2013 · MAE | 10.35 -> 9.18 | Helped | OpticalDR: A Deep Optical Imaging Model for Privacy-Protective Depression Recognition (CVPR 2024 · 14 citations) |
| Add attention map consistency | RAF-DB, Surprise open · AUROC | 0.517 -> 0.885 | Helped | Open-Set Facial Expression Recognition (AAAI 2024 · 13 citations) |
| Add multi-scale facial dynamic modeling (MFD) | BP4D · F1-score (%) | 63.3 -> 64.6 | Helped | Multi-scale Dynamic and Hierarchical Relationship Modeling for Facial Action Units Recognition (CVPR 2024 · 13 citations) |
| Add SSAN joint optimization without distribution loss | CROHME 2014 · ExpRate | 51.48 -> 56.69 | Helped | SSAN: A Symbol Spatial-Aware Network for Handwritten Mathematical Expression Recognition (AAAI 2025 · 7 citations) |
| Add GTVG | SMIC-HS, three-class · UF1 | 0.6105 -> 0.8855 | Helped | Dynamic Stereotype Theory Induced Micro-expression Recognition with Oriented Deformation (CVPR 2025 · 4 citations) |
| Add SAOD to the Vision Transformer baseline | Oulu-CASIA · accuracy | 81.21 -> 82.71 | Helped | Hypergraph-Guided Disentangled Spectrum Transformer Networks for Near-Infrared Facial Expression Recognition (AAAI 2024 · 3 citations) |
| Remove pixel-wise and perceptual losses | DISFA test, target removal and addition · MSE | 0.288 -> 0.388 | Hurt | AUEditNet: Dual-Branch Facial Action Unit Intensity Manipulation with Implicit Disentanglement (CVPR 2024 · 3 citations) |
| replace SDPA with CSA in the cross-attention module | RAF-DB and FERPlus · accuracy | 91.95 -> 92.14; 90.93 -> 91.02 | Helped | QCS:Feature Refining from Quadruplet Cross Similarity for Facial Expression Recognition (AAAI 2025 · 1 citation) |
What reviewers will expect
Datasets and baselines named in the 10 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- RAF-DB · 3 of 10 papers
- FERPlus · 2 of 10 papers
- AffectNet · 2 of 10 papers
- DISFA · 2 of 10 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 17 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 3 papers
- Beijing University of Posts and Telecommunications · 2 papers
Use it
Download BibTeX of 31 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Margin-based face recognition losses (124 papers)
- Supervised person re-identification (108 papers)
- Cross-modality and occlusion re-identification (82 papers)
- Fair and quality-aware face recognition (66 papers)
Elsewhere in the topic
- Multilingual language representation pretraining (507 papers)
- Contrastive representation theory (495 papers)
- Dense features and foundation priors (482 papers)
- Correspondence and causal representation learning (447 papers)
- Data augmentation and generalization bias (409 papers)
- Partial and multi-label learning (345 papers)
- Loss design and classifier theory (273 papers)
- Group-robust classification (251 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).