Research map / Detection and segmentation / Weakly supervised detection and segmentation
End-to-End Semi-Supervised Object Detection With Soft Teacher
Mengde Xu, Zheng Zhang, Han Hu, Jianfeng Wang, Lijuan Wang, Fangyun Wei, Xiang Bai, Zicheng Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| switch from multi-stage to end-to-end pseudo-labeling (ablation) | Faster R-CNN · COCO val2017 (10% labeled, ablation) | mAP: 28.7 -> 30.0 | Improved |
| update teacher by EMA of student (ablation) | E2E · COCO val2017 (10% labeled, ablation) | mAP: 30.0 -> 31.2 | Improved |
| add soft teacher (ablation) | E2E+EMA · COCO val2017 (10% labeled, ablation) | mAP: 31.2 -> 33.6 | Improved |
| add box jittering (ablation) | E2E+EMA+soft teacher · COCO val2017 (10% labeled, ablation) | mAP: 33.6 -> 34.2 | Improved |
| use teacher background score (BG-T) as reliability instead of student (BG-S) (ablation) | soft teacher · COCO val2017 (10% labeled, ablation) | mAP: 25.9 -> 33.6 | Improved |
| use BG-T instead of IoU weighting (ablation) | soft teacher · COCO val2017 (10% labeled, ablation) | mAP: 31.7 -> 33.6 | Improved |
| use BG-T instead of Pred-Diff (ablation) | soft teacher · COCO val2017 (10% labeled, ablation) | mAP: 32.3 -> 33.6 | Improved |
| raise foreground threshold from 0.70 to 0.90 (ablation) | Soft Teacher · COCO val2017 (10% labeled, ablation) | mAP: 29.9 -> 33.6 | Improved |
| raise foreground threshold from 0.90 to 0.95 (ablation) | Soft Teacher · COCO val2017 (10% labeled, ablation) | mAP: 33.6 -> 32.1 | Hurt |
| lower box regression variance threshold from 0.02 to 0.01 (ablation) | Soft Teacher · COCO val2017 (10% labeled, ablation) | mAP: 34.2 -> 32.9 | Hurt |
| increase N_jitter from 10 to 15 (ablation) | box jittering · COCO val2017 (10% labeled, ablation) | mAP: 34.2 -> 34.2 | No clear effect |
| add Soft Teacher with unlabeled2017 to fully labeled training | Faster R-CNN R50 · COCO val2017 (fully labeled + unlabeled2017) | mAP: 40.9 -> 44.5 | Improved |
Closest papers by meaning
- Humble Teachers Teach Better Students for Semi-Supervised Object Detection (CVPR 2021 · similarity 0.94)
- Unbiased Teacher for Semi-Supervised Object Detection (ICLR 2021 · similarity 0.93)
- Instant-Teaching: An End-to-End Semi-Supervised Object Detection Framework (CVPR 2021 · similarity 0.93)
- Diverse Learner: Exploring Diverse Supervision for Semi-Supervised Object Detection (ECCV 2022 · similarity 0.93)
- Dense Teacher: Dense Pseudo-Labels for Semi-Supervised Object Detection (ECCV 2022 · similarity 0.91)
- Semi-supervised 3D Object Detection with PatchTeacher and PillarMix (AAAI 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.
