Research map / Detection and segmentation / Weakly supervised detection and segmentation
Label Matching Semi-Supervised Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace a single fixed confidence threshold with adaptive label-distribution-aware thresholds (ablation) | Mean-teacher pseudo labeling · COCO-standard 1% (ablation setting) | AP50:95 | Helped |
| Stop updating ACT during training (ablation) | LabelMatch · COCO-standard 1% (ablation setting) | AP50:95 | Hurt |
| Use online rather than offline ACT (ablation) | LabelMatch · COCO-standard | AP50:95: 24.6 vs 24.6; 31.6 vs 31.5; 34.6 vs 34.6 | No clear effect |
| Set all candidate pseudo labels as uncertain instead of separating reliable and uncertain labels (ablation) | LabelMatch · COCO-standard 1% (ablation setting) | AP50:95 | Hurt |
| Set too many pseudo labels as reliable (ablation) | LabelMatch · COCO-standard 1% (ablation setting) | AP50:95 | Hurt |
| Use proposal self-assignment for uncertain pseudo labels instead of ignore assignment or IoU-based assignment (ablation) | LabelMatch · COCO-standard 1% (ablation setting) | AP50:95 | Helped |
| Add reliable pseudo label mining to convert high-quality uncertain labels into reliable labels (ablation) | LabelMatch · COCO-standard | AP50:95 | Helped |
| Use RPLM thresholds (0.8, 0.8) rather than other tested threshold pairs (ablation) | LabelMatch · COCO-standard 10% | AP50:95: 34.4; 34.5; 34.6; 34.3 | Mixed |
| Use ACT rather than a fixed threshold in STAC (ablation) | STAC · COCO-standard 1% | AP50:95 | Helped |
| Estimate the unlabeled class distribution using its ground-truth labels rather than the labeled source distribution (ablation) | LabelMatch · Small→Large | AP50: 38.8 to 44.5 | Helped |
Closest papers by meaning
- Humble Teachers Teach Better Students for Semi-Supervised Object Detection (CVPR 2021 · similarity 0.91)
- De-biased Teacher: Rethinking IoU Matching for Semi-supervised Object Detection (AAAI 2023 · similarity 0.91)
- Interactive Self-Training With Mean Teachers for Semi-Supervised Object Detection (CVPR 2021 · similarity 0.91)
- Adapting Object Size Variance and Class Imbalance for Semi-supervised Object Detection (AAAI 2023 · similarity 0.90)
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.90)
- Instant-Teaching: An End-to-End Semi-Supervised Object Detection Framework (CVPR 2021 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.