Research map / Label-efficient and robust learning / Recognition and long-tail
Fair and quality-aware face recognition
Uses quality-adaptive margins, unified thresholds and fairness adapters in face recognition.
66 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
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
7 papers in 2023–24 and 12 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 0.6% to 0.9%.
No longer reported by papers here: LFW (verification accuracy (%), last in 2021).
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 · The MegaFace Benchmark: 1 Million Faces for Recognition at Scale (CVPR 2016 · 978 citations)
- 2017 · Detecting Masked Faces in the Wild With LLE-CNNs (CVPR 2017 · 464 citations)
- 2018 · Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision (CVPR 2018 · 725 citations)
- 2019 · Face Anti-Spoofing: Model Matters, so Does Data (CVPR 2019 · 259 citations)
- 2020 · Data Uncertainty Learning in Face Recognition (CVPR 2020 · 301 citations)
- 2021 · MagFace: A Universal Representation for Face Recognition and Quality Assessment (CVPR 2021 · 617 citations)
- 2022 · AdaFace: Quality Adaptive Margin for Face Recognition (CVPR 2022 · 577 citations)
- 2023 · Implicit Identity Driven Deepfake Face Swapping Detection (CVPR 2023 · 168 citations)
- 2024 · One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · 21 citations)
- 2025 · Mixture-of-Attack-Experts with Class Regularization for Unified Physical-Digital Face Attack Detection (AAAI 2025 · 11 citations)
- 2026 · Efficient, Validation-Free Intrinsic Quality Estimation for Large-Scale Face Recognition Datasets (ICML 2026 · 0 citations)
Earliest papers here
- The MegaFace Benchmark: 1 Million Faces for Recognition at Scale (CVPR 2016 · 978 citations)
- Self-Adaptive Matrix Completion for Heart Rate Estimation From Face Videos Under Realistic Conditions (CVPR 2016 · 352 citations)
- Recurrent Face Aging (CVPR 2016 · 199 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.
LFW · accuracy (higher is better)
4 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 99.71 | Data Uncertainty Learning in Face Recognition (CVPR 2020 · 301 citations) |
| 2025 | 99.79 | AQUAFace: Age-Invariant Quality Adaptive Face Recognition for Unconstrained Selfie vs ID Verification (AAAI 2025 · 2 citations) |
LFW · verification accuracy (%) (higher is better)
4 papers, 2019–2021 · not reported since 2021
| Year | Best so far | Set by |
|---|---|---|
| 2019 | 99.62 | AdaptiveFace: Adaptive Margin and Sampling for Face Recognition (CVPR 2019 · 182 citations) |
| 2020 | 99.8 | RDCFace: Radial Distortion Correction for Face Recognition (CVPR 2020 · 23 citations) |
| 2021 | 99.83 | MagFace: A Universal Representation for Face Recognition and Quality Assessment (CVPR 2021 · 617 citations) |
What has been tried
60 papers here report 628 measured changes (a component added, removed or swapped, compared with a baseline): helped 67%, no clear effect 5%, hurt 23%, 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 SCM-guided generated latent spoof features (LG) | C → I · HTER | 30.79 -> 21.36 | Helped | One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · 21 citations) |
| Add SoftMoE to CLIP | UniAttackData Protocol I · ACER (%) | 0.79 -> 0.66 | Helped | Mixture-of-Attack-Experts with Class Regularization for Unified Physical-Digital Face Attack Detection (AAAI 2025 · 11 citations) |
| Inject public face attributes during patient face synthesis | Generated patient faces · average identity leakage percentage | 71.83% -> 1.27% | Helped | Medical Manifestation-Aware De-Identification (AAAI 2025 · 6 citations) |
| Add Retina Patch and Semantic Attention Head to ViT | Face Avg (CPLFW, CFPFP, CALFW, AGEDB) · accuracy | 90.63 -> 96.66 | Helped | SapiensID: Foundation for Human Recognition (CVPR 2025 · 5 citations) |
| add segmentation-based and patch-based regularizers | Thermalized FAKE vs SEJONG · FID | .5028 -> .1598 | Helped | T-FAKE: Synthesizing Thermal Images for Facial Landmarking (CVPR 2025 · 3 citations) |
| Set margin parameter m to 0.5 rather than 0.2 | VerifyMe easy pairs · performance (Table 5) | 73.42 -> 74.49 | Helped | AQUAFace: Age-Invariant Quality Adaptive Face Recognition for Unconstrained Selfie vs ID Verification (AAAI 2025 · 2 citations) |
| add proxy loss | BTS 3.1 Face Included Treatment, selected experts bK=4 · TAR | 67.32 -> 68.93 | Helped | ProxyFusion: Face Feature Aggregation Through Sparse Experts (NeurIPS 2024 · 1 citation) |
| replace the Register Module with Gaussian noise added to DINOv2 features during adaptation | VFHQ Test · LPIPS | 0.2470 -> 0.2699 | Hurt | Low-Rank Head Avatar Personalization with Registers (NeurIPS 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 7 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- LFW · 3 of 7 papers
- CFP-FP · 2 of 7 papers
- CPLFW · 2 of 7 papers
Baselines
- AdaFace · 2 of 7 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 19 papers; from OpenAlex and the papers' affiliations).
- National Tsing Hua University · 2 papers
Use it
Download BibTeX of 34 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)
- Uncertainty-aware facial expression recognition (41 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).