Research map / Detection and segmentation / Open-set and few-shot detection
Open-set and unknown object detection
Detect unknown objects by modeling latent regions, geometry cues, and class-agnostic prompts.
160 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Open-set and few-shot detection, keyword cluster “images · anomaly · interactive”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Leveling off Several papers from 2025–26 still report on 4 tracked benchmarks, but the best result has barely moved since 2024.
40 papers in 2023–24 and 47 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 6.8% to 8.5%.
What looks settled
- CHAMELEON (Sα): best up to 2024 0.898, best in 2025–26 0.819 from 2 papers, at most 0.2% better, if at all.
- CAMO (Sα): best up to 2024 0.878, best in 2025–26 0.749 from 2 papers, at most 0.2% better, if at all.
- COD10K (Sα): best up to 2024 0.883, best in 2025–26 0.773 from 2 papers, at most 0.2% better, if at all.
- NC4K (Sα): best up to 2024 0.893, best in 2025–26 0.8 from 2 papers, at most 0.2% better, if at all.
No longer reported by papers here: PASCAL VOC 2007 test (mAP, 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 · The Cityscapes Dataset for Semantic Urban Scene Understanding (CVPR 2016 · 12,279 citations)
- 2017 · Perceptual Generative Adversarial Networks for Small Object Detection (CVPR 2017 · 895 citations)
- 2018 · DOTA: A Large-Scale Dataset for Object Detection in Aerial Images (CVPR 2018 · 3,300 citations)
- 2019 · Shifting More Attention to Video Salient Object Detection (CVPR 2019 · 536 citations)
- 2020 · Camouflaged Object Detection (CVPR 2020 · 914 citations)
- 2021 · Simultaneously Localize, Segment and Rank the Camouflaged Objects (CVPR 2021 · 571 citations)
- 2022 · Zoom in and Out: A Mixed-Scale Triplet Network for Camouflaged Object Detection (CVPR 2022 · 480 citations)
- 2023 · PROB: Probabilistic Objectness for Open World Object Detection (CVPR 2023 · 110 citations)
- 2024 · VSCode: General Visual Salient and Camouflaged Object Detection with 2D Prompt Learning (CVPR 2024 · 129 citations)
- 2025 · AeroGen: Enhancing Remote Sensing Object Detection with Diffusion-Driven Data Generation (CVPR 2025 · 33 citations)
- 2026 · Self-Supervised Representation Learning with Joint Embedding Predictive Architecture for Automotive LiDAR Object Detection (AAAI 2026 · 2 citations)
Earliest papers here
- The Cityscapes Dataset for Semantic Urban Scene Understanding (CVPR 2016 · 12,279 citations)
- WIDER FACE: A Face Detection Benchmark (CVPR 2016 · 1,887 citations)
- How Far Are We From Solving Pedestrian Detection? (CVPR 2016 · 431 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.
CHAMELEON · Sα (higher is better)
5 papers, 2020–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 0.869 | Camouflaged Object Detection (CVPR 2020 · 914 citations) |
| 2021 | 0.893 | Simultaneously Localize, Segment and Rank the Camouflaged Objects (CVPR 2021 · 571 citations) |
| 2022 | 0.898 | Detecting Camouflaged Object in Frequency Domain (CVPR 2022 · 316 citations) |
CAMO · Sα (higher is better)
4 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 0.793 | Simultaneously Localize, Segment and Rank the Camouflaged Objects (CVPR 2021 · 571 citations) |
| 2024 | 0.878 | CamoDiffusion: Camouflaged Object Detection via Conditional Diffusion Models (AAAI 2024 · 69 citations) |
COD10K · Sα (higher is better)
4 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 0.793 | Simultaneously Localize, Segment and Rank the Camouflaged Objects (CVPR 2021 · 571 citations) |
| 2024 | 0.883 | CamoDiffusion: Camouflaged Object Detection via Conditional Diffusion Models (AAAI 2024 · 69 citations) |
NC4K · Sα (higher is better)
4 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 0.839 | Simultaneously Localize, Segment and Rank the Camouflaged Objects (CVPR 2021 · 571 citations) |
| 2024 | 0.893 | CamoDiffusion: Camouflaged Object Detection via Conditional Diffusion Models (AAAI 2024 · 69 citations) |
PASCAL VOC 10 + 10 · mAP (higher is better)
4 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 64.58 | Towards Open World Object Detection (CVPR 2021 · 542 citations) |
| 2024 | 72.3 | Exploring Orthogonality in Open World Object Detection (CVPR 2024 · 16 citations) |
PASCAL VOC 2007 test · mAP (higher is better)
4 papers, 2016–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2016 | 75.7 | CRAFT Objects From Images (CVPR 2016 · 134 citations) |
| 2018 | 78.8 | Hybrid Knowledge Routed Modules for Large-scale Object Detection (NeurIPS 2018 · 76 citations) |
| 2019 | 82.5 | Reasoning-RCNN: Unifying Adaptive Global Reasoning Into Large-Scale Object Detection (CVPR 2019 · 110 citations) |
What has been tried
137 papers here report 1,548 measured changes (a component added, removed or swapped, compared with a baseline): helped 73%, no clear effect 4%, 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 |
|---|---|---|---|---|
| Add encoder task-specific prompts | DUTS, NJUD, VT5000, SegV2, CAMO, CAD · Sm | DUTS .908 -> .909; NJUD .932 -> .935; VT5000 .909 -> .912; SegV2 .933 -> .943; CAMO .800 -> .811; CAD .743 -> .736 | Helped | VSCode: General Visual Salient and Camouflaged Object Detection with 2D Prompt Learning (CVPR 2024 · 129 citations) |
| Add ATCN and TTC | COD10K · Sα | 0.867 -> 0.881 | Helped | CamoDiffusion: Camouflaged Object Detection via Conditional Diffusion Models (AAAI 2024 · 69 citations) |
| Add Layout Mask Attention (LMA) | DIOR image generation · FID | 82.11 -> 40.76 | Helped | AeroGen: Enhancing Remote Sensing Object Detection with Diffusion-Driven Data Generation (CVPR 2025 · 33 citations) |
| Add the Data Feeder to AIDE | Novel categories · AP (%) | 10.1 -> 12.0 | Helped | AIDE: An Automatic Data Engine for Object Detection in Autonomous Driving (CVPR 2024 · 30 citations) |
| use TE for training and TIE for inference instead of switchable TE/TIE | ROTX-MP All (train KAIST) · AP | 70.44 -> 57.05 | Hurt | Causal Mode Multiplexer: A Novel Framework for Unbiased Multispectral Pedestrian Detection (CVPR 2024 · 27 citations) |
| Replace hyperbolic distance with cosine distance (c=0) | Hierarchical Split Task 1 · U-Recall | 34.9 -> 32.8 | Hurt | Hyp-OW: Exploiting Hierarchical Structure Learning with Hyperbolic Distance Enhances Open World Object Detection (AAAI 2024 · 16 citations) |
| Add feature orthogonalization | M-OWODB Task 1 · U-Recall | 8.4 -> 18.2 | Helped | Exploring Orthogonality in Open World Object Detection (CVPR 2024 · 16 citations) |
| use object query-guided pseudo-labeling instead of baseline feature averaging | MS COCO Task-1 · U-Recall | 6.17 -> 12.26 | Helped | Semi-supervised Open-World Object Detection (AAAI 2024 · 12 citations) |
| add self-prompting to YOLOv8 | VisDrone · AP | 41.3 -> 42.1 | Helped | Self-Prompting Analogical Reasoning for UAV Object Detection (AAAI 2025 · 11 citations) |
| Add trichromatic AoLP and DoLP to RGB input | RGBP-Car test · AP | 57.6 -> 58.5 | Helped | Exploiting Polarized Material Cues for Robust Car Detection (AAAI 2024 · 11 citations) |
| replace the hybrid evidential loss with CE Bayes-risk loss | VOC-COCO-20 · mAPK | 58.97 -> 48.72 | Hurt | Towards Evidential and Class Separable Open Set Object Detection (AAAI 2024 · 11 citations) |
| Use the joint objectness-class score rather than either score alone | OOD detection benchmarks · OOD detection performance | Helped | YolOOD: Utilizing Object Detection Concepts for Multi-Label Out-of-Distribution Detection (CVPR 2024 · 8 citations) |
What reviewers will expect
Datasets and baselines named in the 32 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- MS-COCO · 8 of 32 papers
- PASCAL VOC · 6 of 32 papers
- OpenImages · 4 of 32 papers
- BDD-100K · 3 of 32 papers
- Objects365 · 3 of 32 papers
- Visual Genome · 2 of 32 papers
- COD10K · 2 of 32 papers
- CAMO · 2 of 32 papers
Baselines
- MSP · 3 of 32 papers
- Energy score · 3 of 32 papers
- VOS · 3 of 32 papers
- Faster R-CNN · 3 of 32 papers
- ORE · 3 of 32 papers
- OW-DETR · 3 of 32 papers
- PROB · 3 of 32 papers
- ODIN · 2 of 32 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 87 papers; from OpenAlex and the papers' affiliations).
- Xidian University · 5 papers
- Chinese Academy of Sciences · 3 papers
- Samsung · 3 papers
- Five AI Ltd · 2 papers
- Tsinghua University · 2 papers
- Beijing University of Posts and Telecommunications · 2 papers
- Zhejiang Lab · 2 papers
- Fudan University · 2 papers
- Huazhong University of Science and Technology · 2 papers
- University of Electronic Science and Technology of China · 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
- Object-level representation learning (219 papers)
- Interactive and referring segmentation (161 papers)
- Amodal and open-world instance segmentation (137 papers)
- Industrial anomaly detection (117 papers)
- Remote sensing segmentation and change (84 papers)
Elsewhere in the topic
- Unsupervised object and part discovery (114 papers)
- Domain-adaptive object detection (103 papers)
- Open-vocabulary object detection (96 papers)
- Open-vocabulary semantic segmentation (93 papers)
- Object-centric learning theory and video (86 papers)
- Few-shot object detection (67 papers)
- Human-object interaction detection (66 papers)
- Few-shot segmentation (64 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).