Research map / Detection and segmentation / Weakly supervised detection and segmentation
MixTeacher: Mining Promising Labels With Mixed Scale Teacher for Semi-Supervised Object Detection
Liang Liu, Boshen Zhang, Jiangning Zhang, Wuhao Zhang, Zhenye Gan, Guanzhong Tian, Wenbing Zhu, Yabiao Wang, Chengjie Wang
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 |
|---|---|---|---|
| use mixed-scale features to generate pseudo labels instead of regular-scale features (ablation) | MixTeacher · COCO partially labeled 10% | mAP: 34.4 -> 36.2 | Improved |
| add down-sampled features while retaining regular-scale teacher supervision (ablation) | semi-supervised baseline · COCO partially labeled 10% | mAP: 33.9 -> 34.7 | Improved |
| add mixed-scale features but keep regular-scale pseudo-label supervision (ablation) | semi-supervised baseline · COCO partially labeled 10% | mAP: 34.7 -> 34.4 | Hurt |
| add Promising Labels Mining to mixed-scale pseudo-label training (ablation) | Mixed Scale Teacher · COCO partially labeled 10% | mAP: 36.2 -> 36.7 | Improved |
| replace GAP-MLP weighted fusion with CONV-ADD (ablation) | Mixed Scale Teacher · COCO partially labeled 10% | mAP: 36.2 -> 35.1 | Hurt |
| replace GAP-MLP weighted fusion with CAT-CONV (ablation) | Mixed Scale Teacher · COCO partially labeled 10% | mAP: 36.2 -> 35.2 | Hurt |
| randomly drop one student path during training (ablation) | Mixed Scale Teacher · COCO partially labeled 10% | mAP: 36.2 -> 36.0 | Hurt |
| test with mixed-scale features instead of regular-scale features (ablation) | MixTeacher · COCO partially labeled 10% | mAP: 36.7 -> 37.5 | Improved |
| test with down-sampled features instead of regular-scale features (ablation) | MixTeacher · COCO partially labeled 10% | mAP: 36.7 -> 33.2 | Hurt |
| set fusion temperature T to 3.0 rather than 0.1 (ablation) | MixTeacher · COCO partially labeled 10% | mAP: 35.9 -> 36.2 | Improved |
| include low-confidence candidates with score threshold τ_l=0.7 rather than 0.6 (ablation) | Promising Labels Mining · COCO partially labeled 10% | mAP: 33.8 -> 36.7 | Improved |
| set promotion threshold δ to 0.1 rather than 0.0 (ablation) | Promising Labels Mining · COCO partially labeled 10% | mAP: 36.2 -> 36.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- PseCo — PseCo: Pseudo Labeling and Consistency Training for Semi-Supervised Object Detection (ECCV 2022) · better on 1 of 2 shared results
- Dense Teacher — Dense Teacher: Dense Pseudo-Labels for Semi-Supervised Object Detection (ECCV 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Humble Teachers Teach Better Students for Semi-Supervised Object Detection (CVPR 2021 · similarity 0.90)
- Diverse Learner: Exploring Diverse Supervision for Semi-Supervised Object Detection (ECCV 2022 · similarity 0.90)
- 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 0.89)
- Scale-Equivalent Distillation for Semi-Supervised Object Detection (CVPR 2022 · similarity 0.89)
- Consistent-Teacher: Towards Reducing Inconsistent Pseudo-Targets in Semi-Supervised Object Detection (CVPR 2023 · 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.
