Research map / Detection and segmentation / Open-set and few-shot detection
Simplifying Source-Free Domain Adaptation for Object Detection: Effective Self-Training Strategies and Performance Insights
Yan Hao, Florent Forest, Olga Fink
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 batch normalization layers to the VGG16 backbone (source-only) (ablation) | Faster-RCNN source-only · Cityscapes->Foggy-Cityscapes | mAP@0.5: 20.0 -> 26.9 | Improved |
| add batch normalization layers to the VGG16 backbone (source-only) (ablation) | Faster-RCNN source-only · SIM10k->Cityscapes | AP on car: 31.0 -> 31.5 | No clear effect |
| adapt batch statistics on target train data (AdaBN) | source model with BN · SIM10k->Cityscapes | AP on car: 31.5 -> 46.9 | Improved |
| adapt batch statistics on target train data (AdaBN) | source model with BN · Cityscapes->Foggy-Cityscapes | mAP@0.5: 26.9 -> 35.1 | Improved |
| enable regression losses on unlabeled target samples (ablation) | SF-UT · Cityscapes->Foggy-Cityscapes | mAP@0.5: 43.3 -> 45.0 | Improved |
| add AdaBN before generating fixed pseudo-labels (ablation) | Fixed SF-FixMatch · Cityscapes->Foggy-Cityscapes | mAP@0.5: 39.0 -> 44.5 | Improved |
| add AdaBN before generating fixed pseudo-labels (ablation) | Fixed SF-PL · Cityscapes->Foggy-Cityscapes | mAP@0.5: 35.6 -> 38.5 | Improved |
| add weak-strong augmentation to AdaBN + Fixed SF-PL (ablation) | AdaBN + Fixed SF · Cityscapes->Foggy-Cityscapes | mAP@0.5: 38.5 -> 44.5 | Improved |
| add weak-strong augmentation to Mean Teacher (giving SF-UT) (ablation) | Mean Teacher · Cityscapes->Foggy-Cityscapes | mAP@0.5: 38.2 -> 45.0 | Improved |
| use fixed pseudo-labels instead of updated ones (ablation) | SF-FixMatch · Cityscapes->Foggy-Cityscapes | mAP@0.5: 36.8 -> 39.0 | Improved |
| add Mosaic augmentation on top of strong augmentation (ablation) | AdaBN + Fixed SF-FM · Cityscapes->Foggy-Cityscapes | mAP@0.5: 44.5 -> 44.6 | No clear effect |
| add Mosaic augmentation on top of strong augmentation (ablation) | AdaBN + Fixed SF-PL · Cityscapes->Foggy-Cityscapes | mAP@0.5: 38.5 -> 40.7 | Improved |
Closest papers by meaning
- Dynamic Retraining-Updating Mean Teacher for Source-Free Object Detection (ECCV 2024 · similarity 0.94)
- Cross-Domain Adaptive Teacher for Object Detection (CVPR 2022 · similarity 0.93)
- Enhancing Source-Free Domain Adaptive Object Detection with Low-confidence Pseudo Label Distillation (ECCV 2024 · similarity 0.93)
- Source-Free Object Detection by Learning To Overlook Domain Style (CVPR 2022 · similarity 0.92)
- Large Self-Supervised Models Bridge the Gap in Domain Adaptive Object Detection (CVPR 2025 · similarity 0.90)
- CAT: Exploiting Inter-Class Dynamics for Domain Adaptive Object Detection (CVPR 2024 · similarity 0.90)
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.
