Research map / Label-efficient and robust learning / Domain adaptation and generalization
Harmonious Teacher for Cross-Domain 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 |
|---|---|---|---|
| Use Harmonious Teacher for cross-domain adaptation | FCOS · Cityscapes→Foggy Cityscapes | mAP: 45.4 -> 50.4 | Helped |
| Use Harmonious Teacher for cross-domain adaptation | FCOS · Cityscapes→BDD100K | mAP: 34.9 -> 40.2 | Helped |
| Use Harmonious Teacher for cross-domain adaptation | FCOS · Sim10K→Cityscapes | AP of car: 59.2 -> 65.5 | Helped |
| Use Harmonious Teacher for cross-domain adaptation | FCOS · KITTI→Cityscapes | AP of car: 60.2 -> 60.3 | Helped |
| Add supervised harmonious loss (ablation) | Baseline FCOS self-training · Cityscapes→Foggy Cityscapes | mAP: 37.3 -> 39.1 | Helped |
| Add unsupervised harmonious loss (ablation) | FCOS with supervised harmonious loss · Cityscapes→Foggy Cityscapes | mAP: 39.1 -> 40.5 | Helped |
| Use the harmony measure to rank and select pseudo labels (ablation) | FCOS with harmonious losses · Cityscapes→Foggy Cityscapes | mAP: 40.5 -> 45.2 | Helped |
| Replace harmony-based pseudo-label ranking with harmony-based weighting of all predictions (ablation) | FCOS with harmonious losses · Cityscapes→Foggy Cityscapes | mAP: 45.2 -> 50.4 | Helped |
| Balance classification and localization scores equally in the harmony measure (β=0.5) (ablation) | Harmonious Teacher · Cityscapes→Foggy Cityscapes | mAP: 48.7 -> 50.4 | Helped |
| Set β to 0.5 rather than 0.25 in the harmony measure (ablation) | Harmonious Teacher · Cityscapes→Foggy Cityscapes | mAP: 49.0 -> 50.4 | Helped |
| Set β to 0.5 rather than 0.75 in the harmony measure (ablation) | Harmonious Teacher · Cityscapes→Foggy Cityscapes | mAP: 50.0 -> 50.4 | Helped |
| Set β to 0.5 rather than 1.0 in the harmony measure (ablation) | Harmonious Teacher · Cityscapes→Foggy Cityscapes | mAP: 49.8 -> 50.4 | Helped |
Closest papers by meaning
- Unbiased Mean Teacher for Cross-Domain Object Detection (CVPR 2021 · similarity 0.91)
- Cross-Domain Adaptive Teacher for Object Detection (CVPR 2022 · similarity 0.91)
- Cross Domain Object Detection by Target-Perceived Dual Branch Distillation (CVPR 2022 · similarity 0.89)
- Mean Teacher DETR with Masked Feature Alignment: A Robust Domain Adaptive Detection Transformer Framework (AAAI 2024 · similarity 0.87)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · similarity 0.87)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.87)
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.