Research map / Detection and segmentation / CNN features
Pedestrian and crowd detection
Detect people in crowds with specialized losses and datasets.
48 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2024) · an approach in CNN features, keyword cluster “detectors · detr · search”
State History Benchmarks Tried Baselines Preprints Who Export
Where it stands
Quiet None of the tracked benchmarks has a report from 2025–26.
6 papers in 2023–24 and 0 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 1.8% to 0.0%.
Open questions the data points to
- Papers report OIM‡ on PRW (mAP (%)) anywhere from 21.3 to 36.9: comparisons against it depend on whose number you take.
- Papers report OIM‡ on PRW (top-1 (%)) anywhere from 49.9 to 75.7: comparisons against it depend on whose number you take.
No longer reported by papers here: PRW (mAP, last in 2022); PRW (top-1, last in 2022); CUHK-SYSU (mAP, last in 2021); COCO test-dev (AP, last in 2023); City Persons reasonable (MR−2, last in 2023); CrowdHuman val (AP, last in 2023).
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 · Traffic-Sign Detection and Classification in the Wild (CVPR 2016 · 1,007 citations)
- 2017 · CityPersons: A Diverse Dataset for Pedestrian Detection (CVPR 2017 · 1,009 citations)
- 2018 · Cascaded Pyramid Network for Multi-Person Pose Estimation (CVPR 2018 · 1,874 citations)
- 2019 · Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training (CVPR 2019 · 432 citations)
- 2020 · NMS by Representative Region: Towards Crowded Pedestrian Detection by Proposal Pairing (CVPR 2020 · 195 citations)
- 2021 · Anchor-Free Person Search (CVPR 2021 · 130 citations)
- 2022 · Cascade Transformers for End-to-End Person Search (CVPR 2022 · 89 citations)
- 2023 · Localized Semantic Feature Mixers for Efficient Pedestrian Detection in Autonomous Driving (CVPR 2023 · 45 citations)
- 2024 · GigaHumanDet: Exploring Full-Body Detection on Gigapixel-Level Images (AAAI 2024 · 4 citations)
Earliest papers here
- Traffic-Sign Detection and Classification in the Wild (CVPR 2016 · 1,007 citations)
- End-To-End People Detection in Crowded Scenes (CVPR 2016 · 551 citations)
- Semantic Channels for Fast Pedestrian Detection (CVPR 2016 · 72 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.
PRW · mAP (higher is better)
7 papers, 2019–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2019 | 37.1 | Query-Guided End-To-End Person Search (CVPR 2019 · 152 citations) |
| 2020 | 47.2 | Instance Guided Proposal Network for Person Search (CVPR 2020 · 119 citations) |
| 2021 | 47.6 | Sequential End-to-end Network for Efficient Person Search (AAAI 2021 · 111 citations) |
| 2022 | 53.3 | Cascade Transformers for End-to-End Person Search (CVPR 2022 · 89 citations) |
PRW · top-1 (higher is better)
6 papers, 2019–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2019 | 76.7 | Query-Guided End-To-End Person Search (CVPR 2019 · 152 citations) |
| 2020 | 87 | Instance Guided Proposal Network for Person Search (CVPR 2020 · 119 citations) |
| 2021 | 87.6 | Sequential End-to-end Network for Efficient Person Search (AAAI 2021 · 111 citations) |
CUHK-SYSU · mAP (higher is better)
5 papers, 2020–2021 · not reported since 2021
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 93.9 | TCTS: A Task-Consistent Two-Stage Framework for Person Search (CVPR 2020 · 141 citations) |
| 2021 | 94.8 | Sequential End-to-end Network for Efficient Person Search (AAAI 2021 · 111 citations) |
COCO test-dev · AP (higher is better)
4 papers, 2018–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 73 | Cascaded Pyramid Network for Multi-Person Pose Estimation (CVPR 2018 · 1,874 citations) |
City Persons reasonable · MR−2 (lower is better)
4 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 8.32 | PedHunter: Occlusion Robust Pedestrian Detector in Crowded Scenes (AAAI 2020 · 113 citations) |
CrowdHuman val · AP (higher is better)
4 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 89.6 | Beta R-CNN: Looking into Pedestrian Detection from Another Perspective (NeurIPS 2020) |
| 2022 | 92.5 | Progressive End-to-End Object Detection in Crowded Scenes (CVPR 2022 · 86 citations) |
What has been tried
48 papers here report 529 measured changes (a component added, removed or swapped, compared with a baseline): helped 78%, no clear effect 5%, hurt 14%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add instance-guided learning | PANDA · AP50 | 79.1 -> 79.6 | Helped | GigaHumanDet: Exploring Full-Body Detection on Gigapixel-Level Images (AAAI 2024 · 4 citations) |
| add the part-body association loss and prediction head | BodyHands · Joint AP | 78.07 -> 85.98 | Helped | PBADet: A One-Stage Anchor-Free Approach for Part-Body Association (ICLR 2024 · 0 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 |
|---|---|---|---|---|
| OIM‡ | PRW | mAP (%) | 21.3 – 36.9 | 2 papers, e.g. AAAI 2020 and CVPR 2019 |
| OIM‡ | PRW | top-1 (%) | 49.9 – 75.7 | 2 papers, e.g. AAAI 2020 and CVPR 2019 |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 6 papers; from OpenAlex and the papers' affiliations).
- Shanghai Jiao Tong University · 2 papers
Use it
Download BibTeX of 22 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
- Region-based and query-based detectors (183 papers)
- Box regression and loss design (94 papers)
- Efficient DETR query design (85 papers)
Elsewhere in the topic
- Context modeling for semantic segmentation (241 papers)
- Local features and context in CNNs (170 papers)
- Context for hard detection conditions (133 papers)
- Domain-specific segmentation and detection (118 papers)
- Convolution-transformer backbones (101 papers)
- Boundary-aware instance segmentation (83 papers)
- Plain and high-resolution ViTs (69 papers)
- Arbitrary-shape scene text detection (61 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).