Research map / Detection and segmentation / Weakly supervised detection and segmentation
Mind the Gap: Polishing Pseudo Labels for Accurate Semi-supervised Object Detection
Read the paperCheck 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 |
|---|---|---|---|
| Add dual pseudo-label polishing: category and bounding-box polishing (ablation) | End-to-end SSOD · PASCAL VOC | AP50:95: 48.70 -> 52.40 | Helped |
| Add bounding-box polishing (ablation) | End-to-end SSOD · PASCAL VOC | AP50:95: 48.70 -> 52.10 | Helped |
| Add category polishing (ablation) | End-to-end SSOD · PASCAL VOC | AP50:95: 48.70 -> 51.40 | Helped |
| Use GIoU instead of L1 loss for box polishing (ablation) | Dual pseudo-label polishing · PASCAL VOC | AP50:95: 51.80 -> 52.40 | Helped |
| Disentangle polished category and box pseudo labels (ablation) | Dual pseudo-label polishing · PASCAL VOC | AP50:95: 52.20 -> 52.40 | Helped |
| Add scale jitter augmentation (ablation) | Ours · PASCAL VOC07 test | AP50: 82.50 -> 84.90 | Helped |
| Set θreg to 0.15 rather than 0.2 (ablation) | Dual pseudo-label polishing · PASCAL VOC | AP50: 82.50 -> 82.20 | Hurt |
| Set θreg to 0.25 rather than 0.2 (ablation) | Dual pseudo-label polishing · PASCAL VOC | AP50:95: 52.40 -> 52.10 | Hurt |
| Set γ to 0.02 rather than 0.06 (ablation) | Bounding-box polishing network · PASCAL VOC | AP50:95: 52.40 -> 51.70 | Hurt |
| Set γ to 0.14 rather than 0.06 (ablation) | Bounding-box polishing network · PASCAL VOC | AP50: 82.50 -> 82.30 | Hurt |
| Set η to 0.3 rather than 0.5 (ablation) | Candidate pseudo-label selection · PASCAL VOC | AP50:95: 52.40 -> 52.30 | Hurt |
| Set η to 0.7 rather than 0.5 (ablation) | Candidate pseudo-label selection · PASCAL VOC | AP50:95: 52.40 -> 52.00 | Hurt |
Closest papers by meaning
- Adapting Object Size Variance and Class Imbalance for Semi-supervised Object Detection (AAAI 2023 · similarity 0.92)
- Rethinking Pseudo Labels for Semi-supervised Object Detection (AAAI 2022 · similarity 0.92)
- Consistent-Teacher: Towards Reducing Inconsistent Pseudo-Targets in Semi-Supervised Object Detection (CVPR 2023 · similarity 0.91)
- Semi-DETR: Semi-Supervised Object Detection With Detection Transformers (CVPR 2023 · similarity 0.91)
- Multi-clue Consistency Learning to Bridge Gaps Between General and Oriented Object in Semi-supervised Detection (AAAI 2025 · similarity 0.89)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.88)
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.