Research map / Detection and segmentation
Weakly supervised detection and segmentation: research map
791 papers on Weakly supervised detection and segmentation in Detection and segmentation, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 7 approaches. The busiest year so far is 2022.
What it is
Drawing boxes and masks by hand is costly. Weakly supervised methods train detectors and segmenters from cheaper labels, such as image-level tags; semi-supervised methods combine a few full labels with many unlabeled images.
Terms you will see on this page:
- Weak supervision
- training with labels coarser than the target output, such as tags instead of masks.
- Class activation map (CAM)
- a heat map of the image regions that drove a classifier's decision.
- Pseudo-labels
- predicted boxes or masks used as training labels.
Start reading with the most cited papers here:
- Learning to Detect Salient Objects With Image-Level Supervision (CVPR 2017 · 1,320 citations)
- ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation (CVPR 2016 · 1,154 citations)
Within Detection and segmentation, its share shrank from 9.2% in 2023–24 to 6.0% in 2025–26 (115 → 72 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Weakly supervised detection and segmentation 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):
- Weakly supervised object localization rising: 6% → 12% of the topic (7 → 9 papers)
- Weakly supervised semantic segmentation rising: 23% → 28% of the topic (26 → 20 papers)
- CAM refinement and prototypes rising: 14% → 18% of the topic (16 → 13 papers)
- Weakly supervised detection falling: 21% → 10% of the topic (24 → 7 papers)
- Consistency regularization for segmentation falling: 10% → 6% of the topic (11 → 4 papers)
Approaches and key papers
Weakly supervised segmentation and localization · 630 papers
Keywords: weakly supervised, supervised semantic, image level
Approaches in this cluster:
- Weakly supervised semantic segmentation (253 papers)
Segment from image-level labels using affinity learning, cue mining, and web data. - Weakly supervised detection (172 papers)
Train detectors from points, videos, or image labels, with weak-to-full frameworks. - Weakly supervised object localization (119 papers)
Improve class activation maps and erasing strategies to localize whole objects. - CAM refinement and prototypes (86 papers)
Refine class activation maps with prototypes, SAM, and co-training for segmentation.
Most cited and most cited since 2024:
- Learning to Detect Salient Objects With Image-Level Supervision (CVPR 2017 · 1,320 citations)
- ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation (CVPR 2016 · 1,154 citations)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · 80 citations)
- Frozen CLIP: A Strong Backbone for Weakly Supervised Semantic Segmentation (CVPR 2024 · 54 citations)
Semi-supervised detection and segmentation · 161 papers
Keywords: semi supervised, pseudo, unlabeled
Approaches in this cluster:
- Pseudo-labeling for semi-supervised detection (119 papers)
Train detectors on unlabeled images with teacher-student pseudo labels and distillation. - Consistency regularization for segmentation (25 papers)
Use weak-to-strong and multi-scale consistency for semi-supervised semantic segmentation. - Semi-supervised instance segmentation (17 papers)
Segment instances with mixed point, box, and unlabeled data via self-training.
Most cited and most cited since 2024:
- Revisiting Weak-to-Strong Consistency in Semi-Supervised Semantic Segmentation (CVPR 2023 · 565 citations)
- Data Distillation: Towards Omni-Supervised Learning (CVPR 2018 · 437 citations)
- CorrMatch: Label Propagation via Correlation Matching for Semi-Supervised Semantic Segmentation (CVPR 2024 · 94 citations)
- RankMatch: Exploring the Better Consistency Regularization for Semi-supervised Semantic Segmentation (CVPR 2024 · 63 citations)
Related topics in Detection and segmentation
- Tracking (575)
- Open-set and few-shot detection (2,864)
- 3D detection and driving (761)
- CNN features (2,941)
