Research map / Detection and segmentation / Weakly supervised detection and segmentation
Semi-DETR: Semi-Supervised Object Detection With Detection Transformers
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 |
|---|---|---|---|
| Use Semi-DETR instead of the DINO SSOD baseline | DINO · COCO-Partial 1%, 5%, and 10% | mAP: +2.10, +2.10, +1.90 mAP | Helped |
| Add Stage-wise Hybrid Matching (ablation) | DINO SSOD · COCO-Partial 10% | mAP: 41.6 -> 42.7 | Helped |
| Add Cross-view Query Consistency (ablation) | Semi-DETR with SHM · COCO-Partial 10% | mAP: 42.7 -> 43.1 | Helped |
| Add Cost-based Pseudo Label Mining (ablation) | Semi-DETR with SHM and CQC · COCO-Partial 10% | mAP: 43.1 -> 43.5 | Helped |
| Replace the proposed ranking-based assignment with Max-IoU (ablation) | SHM first stage · COCO-Partial 10% | mAP: 42.8 -> 11.4 | Hurt |
| Replace the proposed ranking-based assignment with ATSS (ablation) | SHM first stage · COCO-Partial 10% | mAP: 42.8 -> 18.7 | Hurt |
| Replace the proposed assignment with SimOTA (ablation) | SHM first stage · COCO-Partial 10% | mAP: 42.8 -> 42.5 | No clear effect |
| Replace cross-view queries with single-view queries (ablation) | Cross-view Query Consistency · COCO-Partial 10% | mAP: 43.5 -> 43.0 | Hurt |
| Remove RoI features from consistency queries (ablation) | Cross-view Query Consistency · COCO-Partial 10% | mAP: 43.5 -> 42.7 | Hurt |
| Use Cost-based GMM filtering instead of fixed-threshold filtering for consistency labels (ablation) | Consistency pseudo-label mining · COCO-Partial 10% | mAP: +0.8 mAP | Helped |
| Also use CPM-mined labels for classification and regression (ablation) | Semi-DETR · COCO-Partial 10% | mAP: 43.5 -> 42.4 | Hurt |
| Extend one-to-many training to 120k iterations (ablation) | Stage-wise Hybrid Matching · COCO-Partial 10% | mAP: 43.5 -> 44.0 | Helped |
Closest papers by meaning
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.95)
- Mind the Gap: Polishing Pseudo Labels for Accurate Semi-supervised Object Detection (AAAI 2023 · similarity 0.91)
- Consistent-Teacher: Towards Reducing Inconsistent Pseudo-Targets in Semi-Supervised Object Detection (CVPR 2023 · similarity 0.91)
- Unbiased Teacher v2: Semi-Supervised Object Detection for Anchor-Free and Anchor-Based Detectors (CVPR 2022 · similarity 0.91)
- Multi-clue Consistency Learning to Bridge Gaps Between General and Oriented Object in Semi-supervised Detection (AAAI 2025 · similarity 0.90)
- Relational Matching for Weakly Semi-Supervised Oriented 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.