Research map / Detection and segmentation / Weakly supervised detection and segmentation
Unbiased Teacher v2: Semi-Supervised Object Detection for Anchor-Free and Anchor-Based Detectors
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 |
|---|---|---|---|
| use Listen2Student boundary-wise regression loss instead of no unsupervised regression loss (ablation) | Faster R-CNN · COCO-standard | AP55: 29.71 -> 30.78 | Helped |
| use Listen2Student boundary-wise regression loss instead of no unsupervised regression loss (ablation) | Faster R-CNN · COCO-standard | AP95: 0.35 -> 0.58 | Helped |
| replace Listen2Student with confidence thresholding for regression pseudo-labels (ablation) | Faster R-CNN · COCO-standard | AP95: 0.35 -> 0.32 | Hurt |
| replace Listen2Student with confidence thresholding for regression pseudo-labels (ablation) | Faster R-CNN · COCO-standard | AP55: 29.71 -> 30.60 | Helped |
| select FCOS pseudo-boxes by classification score rather than combined box score (ablation) | FCOS · COCO-standard semi-supervised | mAP: 15.12 -> 17.79 | Helped |
| use center-sampling rather than standard label assignment (ablation) | FCOS · COCO-standard semi-supervised | mAP: 17.79 -> 14.96 | Hurt |
| use center-sampling rather than standard label assignment (ablation) | FCOS · fully supervised | mAP: 37.10 -> 38.10 | Helped |
| apply Unbiased Teacher to FCOS rather than Faster R-CNN (ablation) | Unbiased Teacher · COCO-standard 0.5% | mAP: 14.36 -> 10.27 | Hurt |
| apply Unbiased Teacher to FCOS rather than Faster R-CNN (ablation) | Unbiased Teacher · COCO-standard 10% | mAP: 29.56 -> 28.18 | No clear effect |
| use Unbiased Teacher v2 with FCOS instead of Unbiased Teacher | FCOS-ResNet50 · COCO-standard 0.5% | mAP: 10.27 -> 16.25 | Helped |
| use Unbiased Teacher v2 with Faster R-CNN instead of SoftTeacher | Faster R-CNN-ResNet50 · COCO-standard 10% | mAP: 34.04 -> 35.02 | Helped |
| add COCO20cls to the unlabeled set | Unbiased Teacher v2 · VOC07-test | AP 50:95: 56.87 -> 58.08 | Helped |
Closest papers by meaning
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.93)
- Humble Teachers Teach Better Students for Semi-Supervised Object Detection (CVPR 2021 · similarity 0.91)
- Semi-DETR: Semi-Supervised Object Detection With Detection Transformers (CVPR 2023 · similarity 0.91)
- Instant-Teaching: An End-to-End Semi-Supervised Object Detection Framework (CVPR 2021 · similarity 0.90)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.90)
- LEOD: Label-Efficient Object Detection for Event Cameras (CVPR 2024 · 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.