Research map / Detection and segmentation / Open-set and few-shot detection
Masked Retraining Teacher-Student Framework for Domain Adaptive Object Detection
Zijing Zhao, Sitong Wei, Qingchao Chen, Dehui Li, Yifan Yang, Yuxin Peng, Yang 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 |
|---|---|---|---|
| add MAE branch on target feature maps (ablation) | adaptive teacher-student baseline · Cityscapes to Foggy Cityscapes (0.02) | mAP: 44.9 -> 48.3 | Improved |
| add selective retraining (ablation) | adaptive teacher-student baseline · Cityscapes to Foggy Cityscapes (0.02) | mAP: 44.9 -> 48.1 | Improved |
| add dynamic threshold for pseudo labels (ablation) | adaptive teacher-student baseline · Cityscapes to Foggy Cityscapes (0.02) | mAP: 44.9 -> 45.5 | Improved |
| combine all three: MAE, retraining and dynamic threshold (ablation) | adaptive teacher-student baseline · Cityscapes to Foggy Cityscapes (0.02) | mAP: 44.9 -> 51.2 | Improved |
| apply MAE on target images only rather than source images (ablation) | MAE branch · Cityscapes to Foggy Cityscapes (0.02) | mAP: 44.5 -> 48.3 | Improved |
| use MAE as a jointly trained branch rather than pretrain-finetune (ablation) | MRT · Cityscapes to Foggy Cityscapes (0.02) | mAP: 42.7 -> 48.3 | Improved |
| use mask ratio 0.8 rather than 0.5 or 0.9 (ablation) | MAE branch · Cityscapes to Foggy Cityscapes (0.02) | mAP: 47.1 / 46.8 -> 48.3 | Improved |
| decay lambda_mask (hard drop of MAE) instead of keeping it fixed (ablation) | MAE branch · Cityscapes to Foggy Cityscapes (0.02) | mAP: 47.9 -> 48.3 | Improved |
| re-initialize student with MAE-refined weights instead of random or source-only weights (ablation) | selective retraining · Cityscapes to Foggy Cityscapes (0.02) | mAP: 48.1 -> 51.2 | Improved |
| re-initialize student with random weights (ablation) | selective retraining · Cityscapes to Foggy Cityscapes (0.02) | mAP: 44.9 -> 45.5 (vs. 51.2 with MAE weights) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MTTrans — MTTrans: Cross-Domain Object Detection with Mean Teacher Transformer (ECCV 2022) · better on 3 of 3 shared results
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
- Large Self-Supervised Models Bridge the Gap in Domain Adaptive Object Detection (CVPR 2025 · similarity 0.92)
- DA-Ada: Learning Domain-Aware Adapter for Domain Adaptive Object Detection (NeurIPS 2024 · similarity 0.91)
- Mean Teacher DETR with Masked Feature Alignment: A Robust Domain Adaptive Detection Transformer Framework (AAAI 2024 · similarity 0.89)
- Triple Feature Disentanglement for One-Stage Adaptive Object Detection (AAAI 2024 · similarity 0.89)
- Non-exemplar Domain Incremental Object Detection via Learning Domain Bias (AAAI 2024 · similarity 0.88)
- CAT: Exploiting Inter-Class Dynamics for Domain Adaptive Object Detection (CVPR 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.
