Research map / Label-efficient and robust learning / Recognition and long-tail
Cross-modality and occlusion re-identification
Handles infrared, occluded and training-free matching for person re-identification.
82 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · 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 3 of 6 tracked benchmarks.
20 papers in 2023–24 and 25 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 1.6% to 1.9%.
Open questions the data points to
- Market-1501 (mAP): best went from 91.4 (up to 2024) to 94.9 in 2025 — From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations)
- SYSU-MM01 all-search (mAP): best went from 68.56 (up to 2024) to 76.44 in 2025 — From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations)
- Market1501 (Rank-1): best went from 96.1 (up to 2024) to 97.3 in 2025 — From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations)
No longer reported by papers here: DukeMTMC-reID (mAP, last in 2024); SYSU-MM01 all-search single-shot (Rank-1, last in 2024); SYSU-MM01 all-search single-shot (mAP, last in 2024).
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
- 2017 · Beyond Triplet Loss: A Deep Quadruplet Network for Person Re-Identification (CVPR 2017 · 1,196 citations)
- 2018 · Person Transfer GAN to Bridge Domain Gap for Person Re-Identification (CVPR 2018 · 2,058 citations)
- 2019 · Learning to Reduce Dual-Level Discrepancy for Infrared-Visible Person Re-Identification (CVPR 2019 · 512 citations)
- 2020 · Infrared-Visible Cross-Modal Person Re-Identification with an X Modality (AAAI 2020 · 440 citations)
- 2021 · Discover Cross-Modality Nuances for Visible-Infrared Person Re-Identification (CVPR 2021 · 316 citations)
- 2022 · FMCNet: Feature-Level Modality Compensation for Visible-Infrared Person Re-Identification (CVPR 2022 · 282 citations)
- 2023 · Diverse Embedding Expansion Network and Low-Light Cross-Modality Benchmark for Visible-Infrared Person Re-Identification (CVPR 2023 · 317 citations)
- 2024 · Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · 76 citations)
- 2025 · From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations)
- 2026 · X-ReID: Multi-granularity Information Interaction for Video-Based Visible-Infrared Person Re-Identification (AAAI 2026 · 4 citations)
Earliest papers here
- Beyond Triplet Loss: A Deep Quadruplet Network for Person Re-Identification (CVPR 2017 · 1,196 citations)
- Learning Deep Context-Aware Features Over Body and Latent Parts for Person Re-Identification (CVPR 2017 · 704 citations)
- A Multi-Task Deep Network for Person Re-Identification (AAAI 2017 · 203 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.
Market-1501 · mAP (higher is better)
18 papers, 2018–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 76.62 | Dual Attention Matching Network for Context-Aware Feature Sequence Based Person Re-Identification (CVPR 2018 · 510 citations) |
| 2019 | 87.1 | Learning Disentangled Representation for Robust Person Re-identification (NeurIPS 2019 · 36 citations) |
| 2020 | 88.9 | Relation Network for Person Re-Identification (AAAI 2020 · 126 citations) |
| 2023 | 90.4 | DC-Former: Diverse and Compact Transformer for Person Re-identification (AAAI 2023 · 72 citations) |
| 2024 | 91.4 | Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · 16 citations) |
| 2025 | 94.9 | From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations) |
DukeMTMC-reID · mAP (higher is better)
11 papers, 2018–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 64.58 | Dual Attention Matching Network for Context-Aware Feature Sequence Based Person Re-Identification (CVPR 2018 · 510 citations) |
| 2019 | 79.5 | Learning Disentangled Representation for Robust Person Re-identification (NeurIPS 2019 · 36 citations) |
| 2023 | 86.8 | Adaptive Sparse Pairwise Loss for Object Re-Identification (CVPR 2023 · 83 citations) |
SYSU-MM01 all-search · mAP (higher is better)
10 papers, 2022–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 62.51 | FMCNet: Feature-Level Modality Compensation for Visible-Infrared Person Re-Identification (CVPR 2022 · 282 citations) |
| 2023 | 68.56 | Cross-Modality Earth Mover’s Distance for Visible Thermal Person Re-identification (AAAI 2023 · 3 citations) |
| 2025 | 76.44 | From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations) |
Market1501 · Rank-1 (higher is better)
9 papers, 2018–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 87.65 | Pose Transferrable Person Re-Identification (CVPR 2018 · 416 citations) |
| 2019 | 93.9 | Learning Incremental Triplet Margin for Person Re-Identification (AAAI 2019 · 49 citations) |
| 2020 | 96.1 | Semantics-Aligned Representation Learning for Person Re-Identification (AAAI 2020 · 138 citations) |
| 2025 | 97.3 | From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations) |
SYSU-MM01 all-search single-shot · Rank-1 (higher is better)
9 papers, 2020–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 61.6 | Cross-Modality Person Re-Identification With Shared-Specific Feature Transfer (CVPR 2020 · 431 citations) |
| 2021 | 70.58 | Discover Cross-Modality Nuances for Visible-Infrared Person Re-Identification (CVPR 2021 · 316 citations) |
| 2023 | 77.78 | PartMix: Regularization Strategy To Learn Part Discovery for Visible-Infrared Person Re-Identification (CVPR 2023 · 132 citations) |
| 2024 | 81.42 | Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · 76 citations) |
SYSU-MM01 all-search single-shot · mAP (higher is better)
9 papers, 2020–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 63.2 | Cross-Modality Person Re-Identification With Shared-Specific Feature Transfer (CVPR 2020 · 431 citations) |
| 2021 | 68.24 | Discover Cross-Modality Nuances for Visible-Infrared Person Re-Identification (CVPR 2021 · 316 citations) |
| 2023 | 74.62 | PartMix: Regularization Strategy To Learn Part Discovery for Visible-Infrared Person Re-Identification (CVPR 2023 · 132 citations) |
| 2024 | 79.85 | Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · 76 citations) |
What has been tried
65 papers here report 783 measured changes (a component added, removed or swapped, compared with a baseline): helped 80%, no clear effect 2%, hurt 17%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add IP while retaining TGSA | SYSU-MM01 all-search single-shot · rank-1 | 68.38 -> 76.40 | Helped | Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · 76 citations) |
| add shallow contrastive losses | SYSU-MM01 all search · rank-1 accuracy | 49.55 -> 53.21 | Helped | Shallow-Deep Collaborative Learning for Unsupervised Visible-Infrared Person Re-Identification (CVPR 2024 · 69 citations) |
| Add SLE to the AGW baseline | SYSU-MM01 and RegDB · R-1 / mAP | 69.9 / 66.9 -> 71.7 / 69.4; 85.0 / 79.1 -> 89.6 / 84.8 | Helped | High-Order Structure Based Middle-Feature Learning for Visible-Infrared Person Re-identification (AAAI 2024 · 58 citations) |
| add IPG | Market1501 · mAP | 3.34 -> 52.81 | Helped | From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · 23 citations) |
| Set k to 4 instead of 16 | MSMT17 · mAP | 49.1 -> 47.2 | Hurt | Catalyst for Clustering-Based Unsupervised Object Re-identification: Feature Calibration (AAAI 2024 · 23 citations) |
| add sparse encoder | Occluded-Duke · Rank-1 | 65.2 -> 68.6; 3.4% improvement | Helped | Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · 16 citations) |
| use GAC instead of DBSCAN for group clustering | CSG, RoadGroup, i-LIDS MCTS · mAP | Helped | Unsupervised Group Re-identification via Adaptive Clustering-Driven Progressive Learning (AAAI 2024 · 16 citations) | |
| Add N-ULC to the PGM baseline | SYSU-MM01 Indoor Search · rank-1 | 58.80 -> 61.14 | Helped | Relieving Universal Label Noise for Unsupervised Visible-Infrared Person Re-Identification by Inferring from Neighbors (AAAI 2025 · 12 citations) |
| increase λ from small values toward an appropriate value, then increase it beyond the optimum | CUHK03 · mAP | Mixed | Cheb-GR: Rethinking K-nearest Neighbor Search in Re-ranking for Person Re-identification (CVPR 2025 · 12 citations) | |
| Add biometric contrastive loss to ID loss | LTCC, clothes-changing · top-1 accuracy | 54.6 -> 55.4 | Helped | DIFFER: Disentangling Identity Features via Semantic Cues for Clothes-Changing Person Re-ID (CVPR 2025 · 12 citations) |
| Add ISE and use adaptive losses instead of the stage-I losses | PRCC(CC) · mAP | 43.9 -> 53.6 | Helped | Identity-Clothing Similarity Modeling for Unsupervised Clothing Change Person Re-Identification (CVPR 2025 · 9 citations) |
| add HPCL to the PGM baseline | SYSU-MM01 all-search · Rank-1 | 56.3 -> 62.1 | Helped | Learning Commonality, Divergence and Variety for Unsupervised Visible-Infrared Person Re-identification (NeurIPS 2024 · 9 citations) |
What reviewers will expect
Datasets and baselines named in the 12 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- SYSU-MM01 · 4 of 12 papers
- Market-1501 · 4 of 12 papers
- MSMT17 · 3 of 12 papers
- RegDB · 2 of 12 papers
- LTCC · 2 of 12 papers
- PRCC · 2 of 12 papers
- Celeb-reID-light · 2 of 12 papers
Baselines
- PPLR · 3 of 12 papers
- TransReID · 3 of 12 papers
- SAAI · 2 of 12 papers
- SPCL · 2 of 12 papers
- ICE · 2 of 12 papers
- OTLA · 2 of 12 papers
- GI-ReID · 2 of 12 papers
- CAL · 2 of 12 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 45 papers; from OpenAlex and the papers' affiliations).
- Wuhan University · 5 papers
- Xiamen University · 4 papers
- Sun Yat-sen University · 4 papers
- Key Laboratory of Machine Intelligence and Advanced Computing · 4 papers
- Xidian University · 2 papers
- Guangdong Key Laboratory of Information Security Technology · 2 papers
- Alibaba Group · 2 papers
- National University of Singapore · 2 papers
- Harbin Institute of Technology · 2 papers
Use it
Download BibTeX of 40 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)
- Fair and quality-aware face recognition (66 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).