Research map / Detection and segmentation / Open-set and few-shot detection
Dynamic Retraining-Updating Mean Teacher for Source-Free Object Detection
Trinh Le Ba Khanh, Huy-Hung Nguyen, Long Hoang Pham, Duong Nguyen-Ngoc Tran, Jae Wook Jeon
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 |
|---|---|---|---|
| add Mean Teacher self-training to source-only model (ablation) | Deformable DETR · Cityscapes -> Foggy Cityscapes | mAP: 29.5 -> 37.4 | Improved |
| add masked image consistency (MIC) to MT (ablation) | MT · Cityscapes -> Foggy Cityscapes | mAP: 37.4 -> 39.8 | Improved |
| add Historical Student Loss (ablation) | MT + MIC · Cityscapes -> Foggy Cityscapes | mAP: 39.8 -> 41.3 | Improved |
| add Dynamic Retraining-Updating on top of Historical Student Loss (ablation) | MT + MIC + L_his · Cityscapes -> Foggy Cityscapes | mAP: 41.3 -> 43.6 | Improved |
| add DRU without Historical Student Loss (ablation) | MT + MIC · Cityscapes -> Foggy Cityscapes | mAP: 39.8 -> 40.9 | Improved |
| re-initialize decoder only instead of backbone+encoder (ablation) | DRU student retraining · Cityscapes -> Foggy Cityscapes | mAP: 39.9 -> 43.6 | Improved |
| re-initialize backbone+encoder instead of continuing training (ablation) | DRU student retraining · Cityscapes -> Foggy Cityscapes | mAP: 41.2 -> 39.9 | Hurt |
| dynamic teacher update interval instead of fixed interval (ablation) | DRU · Cityscapes -> Foggy Cityscapes | mAP: 43.4 (i=5) -> 43.6 | Improved |
| vary meta-iteration M from 5 to 2 or 20 (ablation) | DRU · Cityscapes -> Foggy Cityscapes | mAP: 43.6 -> 41.9 (M=2), 41.5 (M=20) | Hurt |
| DRU vs previous SFOD methods | DRU · Cityscapes -> BDD100k | mAP: 31.6 -> 36.6 | Improved |
Closest papers by meaning
- Simplifying Source-Free Domain Adaptation for Object Detection: Effective Self-Training Strategies and Performance Insights (ECCV 2024 · similarity 0.94)
- Enhancing Source-Free Domain Adaptive Object Detection with Low-confidence Pseudo Label Distillation (ECCV 2024 · similarity 0.93)
- Periodically Exchange Teacher-Student for Source-Free Object Detection (ICCV 2023 · similarity 0.93)
- Instance Relation Graph Guided Source-Free Domain Adaptive Object Detection (CVPR 2023 · similarity 0.92)
- CAT: Exploiting Inter-Class Dynamics for Domain Adaptive Object Detection (CVPR 2024 · similarity 0.90)
- Large Self-Supervised Models Bridge the Gap in Domain Adaptive Object Detection (CVPR 2025 · 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.
