Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Domain Adaptation for One-Stage Object Detector Using Offsets to Bounding Box
Jayeon Yoo, Inseop Chung, Nojun Kwak
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 offset-aware conditioning (left offset) instead of global alignment only | FCOS domain-adaptive detector · Cityscapes to Foggy Cityscapes | mAP@0.5: 37.5 -> 42.6 | Improved |
| condition on both left and top offsets instead of left only (ablation) | OADA · Cityscapes to Foggy Cityscapes | mAP@0.5: 42.6 -> 44.8 | Improved |
| add self-training initialized from OADA pre-trained model | OADA (Offset-Left & Top) · Cityscapes to Foggy Cityscapes | mAP@0.5: 44.8 -> 45.4 | Improved |
| add self-training | OADA (Offset-Left & Top) · Sim10k to Cityscapes | mAP@0.5: 56.6 -> 59.2 | Improved |
| add self-training | OADA (Offset-Left & Top) · KITTI to Cityscapes | mAP@0.5: 46.3 -> 47.8 | Improved |
| smooth the offset probability with alpha0=0.2 rather than none (ablation) | OADA (Offset-Left & Top) · Cityscapes to Foggy Cityscapes | mAP@0.5: 42.6 -> 44.8 | Improved |
| use offset-aware conditioning rather than uniform vector (alpha0=1.0) (ablation) | OADA (Offset-Left & Top) · Cityscapes to Foggy Cityscapes | mAP@0.5: 42.6 -> 44.8 | Improved |
| use 3 offset bins instead of 1 (ablation) | OADA (Offset-Left & Top) · Cityscapes to Foggy Cityscapes | mAP@0.5: 42.4 -> 44.8 | Improved |
| increase offset bins from 3 to 5 (ablation) | OADA (Offset-Left & Top) · Cityscapes to Foggy Cityscapes | mAP@0.5: 44.8 -> 41.4 | Hurt |
| use confidence threshold rho=0.3 instead of 0.0 (ablation) | OADA (Offset-Top) · Cityscapes to Foggy Cityscapes | mAP@0.5: 37.0 -> 42.7 | Improved |
| use confidence threshold rho=0.5 instead of 0.0 (ablation) | OADA (Offset-Top) · Sim10k to Cityscapes | mAP@0.5: 48.5 -> 55.7 | Improved |
| use confidence threshold rho=0.5 instead of 0.0 (ablation) | OADA (Offset-Top) · KITTI to Cityscapes | mAP@0.5: 43.8 -> 45.8 | Improved |
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.
- MeGA-CDA — MeGA-CDA: Memory Guided Attention for Category-Aware Unsupervised Domain Adaptive Object Detection (CVPR 2021) · better on 2 of 2 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
- Adapting Object Detectors via Selective Cross-Domain Alignment (CVPR 2019 · similarity 0.92)
- Domain Adaptive Object Detection via Balancing Between Self-Training and Adversarial Learning (TPAMI 2023 · similarity 0.91)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.90)
- Universal Domain Adaptive Object Detection via Dual Probabilistic Alignment (AAAI 2025 · similarity 0.89)
- Mean Teacher DETR with Masked Feature Alignment: A Robust Domain Adaptive Detection Transformer Framework (AAAI 2024 · similarity 0.89)
- Differential Alignment for Domain Adaptive Object Detection (AAAI 2025 · 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.
