Research map / Video / Video segmentation and tracking
Spatial-temporal video person re-identification
Matches people across video using spatiotemporal attention, graphs and keypoint message passing.
41 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Video segmentation and tracking, keyword cluster “prediction · person · pose”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 6 tracked benchmarks.
4 papers in 2023–24 and 4 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 0.9% to 0.7%.
Open questions the data points to
- iLIDS-VID (Rank-1): best went from 94.5 (up to 2024) to 96.7 in 2025 — CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification (AAAI 2025 · 21 citations)
- Papers report Baseline on MARS (mAP) anywhere from 58.1 to 81.2: comparisons against it depend on whose number you take.
What looks settled
- MARS (mAP): best up to 2024 96.6, best in 2025–26 91.5 from 2 papers, at most 0.2% better, if at all.
- MARS (Rank-1): best up to 2024 93, best in 2025–26 93.3 from 2 papers, less than 1% better.
- LS-VID (mAP): best up to 2024 93.4, best in 2025–26 87.3 from 2 papers, at most 0.2% better, if at all.
No longer reported by papers here: PRID-2011 (Rank-1, last in 2022).
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 · Recurrent Convolutional Network for Video-Based Person Re-Identification (CVPR 2016 · 596 citations)
- 2017 · See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 citations)
- 2018 · Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-Identification (CVPR 2018 · 361 citations)
- 2019 · VRSTC: Occlusion-Free Video Person Re-Identification (CVPR 2019 · 245 citations)
- 2020 · Spatial-Temporal Graph Convolutional Network for Video-Based Person Re-Identification (CVPR 2020 · 248 citations)
- 2021 · BiCnet-TKS: Learning Efficient Spatial-Temporal Representation for Video Person Re-Identification (CVPR 2021 · 115 citations)
- 2022 · Clothes-Changing Person Re-Identification With RGB Modality Only (CVPR 2022 · 253 citations)
- 2024 · TF-CLIP: Learning Text-Free CLIP for Video-Based Person Re-identification (AAAI 2024 · 57 citations)
- 2025 · CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification (AAAI 2025 · 21 citations)
- 2026 · Skeletons Speak Louder than Text: A Motion-Aware Pretraining Paradigm for Video-Based Person Re-Identification (AAAI 2026 · 0 citations)
Earliest papers here
- Recurrent Convolutional Network for Video-Based Person Re-Identification (CVPR 2016 · 596 citations)
- Top-Push Video-Based Person Re-Identification (CVPR 2016 · 220 citations)
- See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 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.
MARS · mAP (higher is better)
27 papers, 2017–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 50.7 | See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 citations) |
| 2018 | 65.8 | Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-Identification (CVPR 2018 · 361 citations) |
| 2019 | 82.3 | VRSTC: Occlusion-Free Video Person Re-Identification (CVPR 2019 · 245 citations) |
| 2020 | 85.8 | Learning Multi-Granular Hypergraphs for Video-Based Person Re-Identification (CVPR 2020 · 161 citations) |
| 2021 | 86.7 | Spatial-Temporal Correlation and Topology Learning for Person Re-Identification in Videos (CVPR 2021 · 79 citations) |
| 2024 | 96.6 | SEAS: ShapE-Aligned Supervision for Person Re-Identification (CVPR 2024 · 18 citations) |
MARS · Rank-1 (higher is better)
15 papers, 2017–2025 · small gains only
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 70.6 | See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 citations) |
| 2019 | 88.5 | VRSTC: Occlusion-Free Video Person Re-Identification (CVPR 2019 · 245 citations) |
| 2020 | 89.95 | Spatial-Temporal Graph Convolutional Network for Video-Based Person Re-Identification (CVPR 2020 · 248 citations) |
| 2021 | 91.4 | Spatial-Temporal Correlation and Topology Learning for Person Re-Identification in Videos (CVPR 2021 · 79 citations) |
| 2024 | 93 | TF-CLIP: Learning Text-Free CLIP for Video-Based Person Re-identification (AAAI 2024 · 57 citations) |
| 2025 | 93.3 | CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification (AAAI 2025 · 21 citations) |
iLIDS-VID · Rank-1 (higher is better)
14 papers, 2017–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 55.2 | See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 citations) |
| 2019 | 86.3 | Attribute-Driven Feature Disentangling and Temporal Aggregation for Video Person Re-Identification (CVPR 2019 · 190 citations) |
| 2020 | 87.2 | Appearance and Motion Enhancement for Video-Based Person Re-Identification (AAAI 2020 · 26 citations) |
| 2021 | 90.4 | Watching You: Global-Guided Reciprocal Learning for Video-Based Person Re-Identification (CVPR 2021 · 110 citations) |
| 2022 | 92.5 | Salient-to-Broad Transition for Video Person Re-Identification (CVPR 2022 · 79 citations) |
| 2024 | 94.5 | TF-CLIP: Learning Text-Free CLIP for Video-Based Person Re-identification (AAAI 2024 · 57 citations) |
| 2025 | 96.7 | CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification (AAAI 2025 · 21 citations) |
LS-VID · mAP (higher is better)
7 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 75.1 | BiCnet-TKS: Learning Efficient Spatial-Temporal Representation for Video Person Re-Identification (CVPR 2021 · 115 citations) |
| 2022 | 79.6 | Salient-to-Broad Transition for Video Person Re-Identification (CVPR 2022 · 79 citations) |
| 2024 | 93.4 | SEAS: ShapE-Aligned Supervision for Person Re-Identification (CVPR 2024 · 18 citations) |
PRID-2011 · Rank-1 (higher is better)
7 papers, 2017–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 79.4 | See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 citations) |
| 2019 | 93.9 | Attribute-Driven Feature Disentangling and Temporal Aggregation for Video Person Re-Identification (CVPR 2019 · 190 citations) |
| 2021 | 96.2 | Watching You: Global-Guided Reciprocal Learning for Video-Based Person Re-Identification (CVPR 2021 · 110 citations) |
| 2022 | 96.5 | Salient-to-Broad Transition for Video Person Re-Identification (CVPR 2022 · 79 citations) |
iLIDS-VID · Rank-5 (higher is better)
6 papers, 2017–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 86.5 | See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-Identification (CVPR 2017 · 348 citations) |
| 2019 | 95.5 | VRSTC: Occlusion-Free Video Person Re-Identification (CVPR 2019 · 245 citations) |
| 2024 | 99.3 | Temporal Correlation Vision Transformer for Video Person Re-Identification (AAAI 2024 · 16 citations) |
| 2025 | 99.9 | CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification (AAAI 2025 · 21 citations) |
What has been tried
39 papers here report 479 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 5%, hurt 20%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| replace CLIP-ReID text branch with identity-specific CLIP-Memory | MARS · mAP | 88.1 -> 88.4 | Helped | TF-CLIP: Learning Text-Free CLIP for Video-Based Person Re-identification (AAAI 2024 · 57 citations) |
| add MMC to the CLIP-Memory baseline | LS-VID · mAP | 80.3 -> 82.0 | Helped | CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification (AAAI 2025 · 21 citations) |
| Add PIFu features as a second branch and concatenate with appearance | Market1501 · Rank-1 accuracy | +0.0 | No clear effect | SEAS: ShapE-Aligned Supervision for Person Re-Identification (CVPR 2024 · 18 citations) |
| add TCA and LTA | LS-VID · mAP | 80.5 -> 83.1 | Helped | Temporal Correlation Vision Transformer for Video Person Re-Identification (AAAI 2024 · 16 citations) |
| Add the Adapted Temporal-Spatial stream | AG-VPReID Aerial-to-Ground · Rank-1 | 65.70 -> 71.91 | Helped | AG-VPReID: A Challenging Large-Scale Benchmark for Aerial-Ground Video-based Person Re-Identification (CVPR 2025 · 15 citations) |
| Add bias-less distillation | NTU RGB-AB, same activity, View+ · Rank 1 accuracy | 64.23 -> 69.31 | Helped | Activity-Biometrics: Person Identification from Daily Activities (CVPR 2024 · 7 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| Baseline | MARS | mAP | 58.1 – 81.2 | 4 papers, e.g. CVPR 2018 and CVPR 2021 |
What reviewers will expect
Datasets and baselines named in the 6 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- MARS · 5 of 6 papers
- LS-VID · 5 of 6 papers
- iLIDS-VID · 4 of 6 papers
- Market-1501 · 2 of 6 papers
- MSMT17 · 2 of 6 papers
Baselines
- CAVIT · 5 of 6 papers
- BiCnet-TKS · 4 of 6 papers
- STMN · 4 of 6 papers
- M3D · 3 of 6 papers
- GLTR · 3 of 6 papers
- TCLNet · 3 of 6 papers
- MGH · 3 of 6 papers
- PSTA · 3 of 6 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 8 papers; from OpenAlex and the papers' affiliations).
- Dalian University of Technology · 2 papers
- Wuhan University of Technology · 2 papers
Use it
Download BibTeX of 28 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
- Layered video decomposition and dynamics (140 papers)
- Stochastic video prediction (125 papers)
- Physics-inspired motion prediction (113 papers)
Elsewhere in the topic
- Self-supervised video representation learning (167 papers)
- Video restoration with temporal priors (159 papers)
- Sign language and physiological video analysis (99 papers)
- Weakly supervised video anomaly detection (99 papers)
- Spatio-temporal predictive modeling (91 papers)
- Matching-based video object segmentation (86 papers)
- Learned optical flow estimation (73 papers)
- Learned video compression (68 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).