Research map / Detection and segmentation / Open-set and few-shot detection
Decoupled Adaptation 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 |
|---|---|---|---|
| decouple category and box adaptation from detector training, with category and box adaptors | D-adapt · VOC-to-Clipart | mAP: 42.1 -> 49.0 | Helped |
| decouple category and box adaptation from detector training, with category and box adaptors | D-adapt · VOC-to-Comic2k | mAP: 33.5 -> 40.5 | Helped |
| add CycleGAN source-to-target translation | D-adapt · Cityscapes-to-FoggyCityscapes | mAP: 38.1 -> 41.3 | Helped |
| add confidence weighting to category adaptation (ablation) | D-adapt category adaptor · VOC-to-Clipart | mAP: 33.3 -> 43.5 | Helped |
| add the category adaptor instead of directly using detector target labels (ablation) | D-adapt · VOC-to-Clipart | mAP: 28.0 -> 43.5 | Helped |
| add disparity discrepancy to box adaptation (ablation) | D-adapt box adaptor · VOC-to-Clipart | mIoU_reg: 0.598 -> 0.631 | Helped |
| add the box adaptor (ablation) | D-adapt · VOC-to-Clipart | mAP: 43.5 -> 45.0 | Helped |
| use standard confidence-threshold pseudo-label training instead of the proposed proposal-region strategy (ablation) | D-adapt · VOC-to-Clipart | mAP: 43.5 -> 38.9; 37.3; 35.9; 34.4 | Hurt |
| use ground-truth IoU proposal selection with a higher threshold (ablation) | D-adapt category adaptor · VOC-to-Clipart | mIoU_cls: 36.1 -> 51.4 | Helped |
| share the category and box adaptors' input proposal distribution (ablation) | D-adapt adaptors · VOC-to-Clipart | mIoU_reg: 0.631 -> 0.319; 0.551 | Mixed |
| add the box adaptor (ablation) | D-adapt · VOC-to-Clipart (T=3) | mAP: 47.0 -> 49.1 | Helped |
Closest papers by meaning
- Adapting Object Detectors via Selective Cross-Domain Alignment (CVPR 2019 · similarity 0.92)
- Cross-domain Object Detection through Coarse-to-Fine Feature Adaptation (CVPR 2020 · similarity 0.91)
- SCAN: Cross Domain Object Detection with Semantic Conditioned Adaptation (AAAI 2022 · similarity 0.90)
- DA-Ada: Learning Domain-Aware Adapter for Domain Adaptive Object Detection (NeurIPS 2024 · similarity 0.87)
- Triple Feature Disentanglement for One-Stage Adaptive Object Detection (AAAI 2024 · similarity 0.87)
- CAT: Exploiting Inter-Class Dynamics for Domain Adaptive Object Detection (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.