Research map / Detection and segmentation
Open-set and few-shot detection: research map
2,864 papers on Open-set and few-shot detection in Detection and segmentation, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 6 clusters and 22 approaches. The busiest year so far is 2024.
What it is
Classic detectors only find the classes they were trained on. This topic builds detectors and segmenters that handle new classes: from a text description (open-vocabulary), from a few examples (few-shot), or by flagging unknown objects, often using vision-language models and the Segment Anything Model.
Terms you will see on this page:
- Open-vocabulary detection
- detecting any object named in text, not just training classes.
- SAM
- the Segment Anything Model, which segments objects from points or boxes.
- Human-object interaction (HOI)
- detecting people, objects and the action between them.
Start reading with the most cited papers here:
- The Cityscapes Dataset for Semantic Urban Scene Understanding (CVPR 2016 · 12,279 citations)
- Scene Parsing Through ADE20K Dataset (CVPR 2017 · 3,305 citations)
Within Detection and segmentation, its share held steady from 46.8% in 2023–24 to 46.0% in 2025–26 (588 → 554 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Open-set and few-shot detection in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Open-vocabulary semantic segmentation rising: 5% → 11% of the topic (30 → 60 papers)
- Industrial anomaly detection rising: 6% → 9% of the topic (34 → 51 papers)
- Remote sensing segmentation and change rising: 4% → 7% of the topic (26 → 39 papers)
- Object-level representation learning falling: 10% → 7% of the topic (61 → 38 papers)
- Unsupervised object and part discovery falling: 5% → 2% of the topic (28 → 9 papers)
Papers from 2025–2026 that are already cited most:
- 5%>100%: Breaking Performance Shackles of Full Fine-Tuning on Visual Recognition Tasks (CVPR 2025 · 70 citations)
- SegEarth-OV: Towards Training-Free Open-Vocabulary Segmentation for Remote Sensing Images (CVPR 2025 · 51 citations)
- LLMDet: Learning Strong Open-Vocabulary Object Detectors under the Supervision of Large Language Models (CVPR 2025 · 41 citations)
- Domain Generalized Medical Landmark Detection via Robust Boundary-Aware Pre-Training (AAAI 2025 · 5 citations)
Approaches and key papers
Open-world and specialised segmentation · 1,327 papers
Keywords: images, anomaly, interactive
Approaches in this cluster:
- Object-level representation learning (315 papers)
Learn object-centric recognition features through contrastive pretraining and context debiasing. - Open-set and unknown object detection (248 papers)
Detect unknown objects by modeling latent regions, geometry cues, and class-agnostic prompts. - Interactive and referring segmentation (246 papers)
Segment objects from clicks or language expressions with refinement pipelines. - Amodal and open-world instance segmentation (224 papers)
Segment instances, including occluded ones, without annotations using pseudo labels. - Industrial anomaly detection (171 papers)
Detect anomalies with student-teacher, multi-sensor, and likelihood-based methods. - Remote sensing segmentation and change (123 papers)
Build datasets and models for geospatial segmentation and change detection.
Most cited and most cited since 2024:
- The Cityscapes Dataset for Semantic Urban Scene Understanding (CVPR 2016 · 12,279 citations)
- Scene Parsing Through ADE20K Dataset (CVPR 2017 · 3,305 citations)
- VSCode: General Visual Salient and Camouflaged Object Detection with 2D Prompt Learning (CVPR 2024 · 129 citations)
- Real-IAD: A Real-World Multi-View Dataset for Benchmarking Versatile Industrial Anomaly Detection (CVPR 2024 · 127 citations)
Interactions and relationships · 418 papers
Keywords: human, interaction, hoi
Approaches in this cluster:
- Unsupervised object and part discovery (146 papers)
Discover objects and parts without labels through contrastive reconstruction and background cues. - Human-object interaction detection (111 papers)
Detect human-object interactions with cross-modal, pose-aware, and knowledge-guided models. - Visual relationship and scene graph detection (95 papers)
Detect object relations and generate scene graphs with context and reinforcement learning. - Hand-object and articulated interaction (66 papers)
Model and generate hand-object and articulated object manipulation.
Most cited and most cited since 2024:
- DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks (CVPR 2018 · 1,784 citations)
- Unsupervised Representation Learning by Predicting Image Rotations (ICLR 2018 · 1,506 citations)
- Improving Audio-Visual Segmentation with Bidirectional Generation (AAAI 2024 · 32 citations)
- Error Detection in Egocentric Procedural Task Videos (CVPR 2024 · 31 citations)
Open-vocabulary detection · 399 papers
Keywords: vocabulary, language, text
Approaches in this cluster:
- Open-vocabulary object detection (153 papers)
Detect novel categories by distilling vision-language knowledge into detectors. - Open-vocabulary semantic segmentation (180 papers)
Segment arbitrary categories using CLIP, mask adaptation, and diffusion without dense labels. - Open-vocabulary 3D and world perception (66 papers)
Extend open-vocabulary understanding to 3D instances, open-world detection, and language-guided segmentation.
Most cited and most cited since 2024:
- LVIS: A Dataset for Large Vocabulary Instance Segmentation (CVPR 2019 · 1,281 citations)
- VinVL: Revisiting Visual Representations in Vision-Language Models (CVPR 2021 · 894 citations)
- YOLO-World: Real-Time Open-Vocabulary Object Detection (CVPR 2024 · 721 citations)
- OMG-Seg: Is One Model Good Enough For All Segmentation? (CVPR 2024 · 77 citations)
Few-shot and zero-shot detection · 352 papers
Keywords: zero, shot object, shot segmentation
Approaches in this cluster:
- Few-shot object detection (103 papers)
Detect novel classes from few examples with contrastive proposals and incremental learning. - Few-shot segmentation (110 papers)
Segment novel classes from few support images using pretrained features and aggregation. - Zero-shot segmentation and detection (79 papers)
Recognize unseen classes via generative features and semantic alignment. - Prompted SAM segmentation (60 papers)
Adapt Segment Anything for semantic, point-prompted, and efficient few-shot segmentation.
Most cited and most cited since 2024:
- Grounded Language-Image Pre-Training (CVPR 2022 · 1,055 citations)
- Synthesized Classifiers for Zero-Shot Learning (CVPR 2016 · 713 citations)
- EfficientSAM: Leveraged Masked Image Pretraining for Efficient Segment Anything (CVPR 2024 · 240 citations)
- AnyDoor: Zero-shot Object-level Image Customization (CVPR 2024 · 175 citations)
Domain-adaptive detection · 245 papers
Keywords: domain, adaptation, adaptive object
Approaches in this cluster:
- Domain-adaptive object detection (174 papers)
Adapt detectors across domains with alignment, diversity, and mean-teacher methods. - Domain-adaptive segmentation (32 papers)
Transfer segmentation across domains using pixel-level association and foundation models. - Cross-domain translation and pose (39 papers)
Handle cross-domain tasks through relation-aware teachers, image translation, and decomposition.
Most cited and most cited since 2024:
- Domain Adaptive Faster R-CNN for Object Detection in the Wild (CVPR 2018 · 1,649 citations)
- Strong-Weak Distribution Alignment for Adaptive Object Detection (CVPR 2019 · 824 citations)
- Stronger Fewer & Superior: Harnessing Vision Foundation Models for Domain Generalized Semantic Segmentation (CVPR 2024 · 101 citations)
- Boosting Object Detection with Zero-Shot Day-Night Domain Adaptation (CVPR 2024 · 71 citations)
Object-centric learning · 123 papers
Keywords: object centric, slot, representations
Approaches in this cluster:
- Object-centric learning theory and video (100 papers)
Learn object-centric representations with identifiability, robustness analysis, and video conditioning. - Slot attention methods (23 papers)
Extend slot attention with grounding, invariance, and query optimization for object discovery.
Most cited and most cited since 2024:
- Object-Centric Auto-Encoders and Dummy Anomalies for Abnormal Event Detection in Video (CVPR 2019 · 432 citations)
- Object-Centric Learning with Slot Attention (NeurIPS 2020 · 211 citations)
- Prompt-Driven Dynamic Object-Centric Learning for Single Domain Generalization (CVPR 2024 · 26 citations)
- EAGLE: Eigen Aggregation Learning for Object-Centric Unsupervised Semantic Segmentation (CVPR 2024 · 20 citations)
Related topics in Detection and segmentation
- Tracking (575)
- 3D detection and driving (761)
- CNN features (2,941)
- Weakly supervised detection and segmentation (791)
