Research map / Detection and segmentation / Open-set and few-shot detection
Domain Adaptive Object Detection via Dual-Stream Bilevel-Cycle Optimization
Yannan Chen, Wei Wang, Ruoyu Chen, Wenqiang Wang, Jiancheng Wang, Mingbo Yang, Yaowei Wang, Xiaochun Cao
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 instance consistency distillation to standard self-training (ablation) | Standard ST (FCOS) · Cityscapes -> Foggy Cityscapes | mAP: 41.2 -> 48.3 | Improved |
| add classification cycle stream on top of instance consistency (ablation) | DSBCO · Cityscapes -> Foggy Cityscapes | mAP: 48.3 -> 54.5 | Improved |
| add regression cycle stream (with normalization) on top of instance consistency (ablation) | DSBCO · Cityscapes -> Foggy Cityscapes | mAP: 48.3 -> 50.9 | Improved |
| combine classification and regression cycle streams (ablation) | DSBCO · Cityscapes -> Foggy Cityscapes | mAP: 54.5 -> 64.7 | Improved |
| DSBCO vs HT (FCOS second best) | DSBCO · Cityscapes -> Foggy Cityscapes | mAP: 50.4 -> 64.7 | Improved |
| DSBCO vs HT | DSBCO · Cityscapes -> BDD100K | mAP: 40.2 -> 41.9 | Improved |
| DSBCO vs HT | DSBCO · Sim10K -> Cityscapes | AP of Car: 65.5 -> 68.6 | Improved |
| DSBCO vs MGCAMT (RetinaNet) | DSBCO · Cityscapes -> BDD100K | mAP: 44.8 vs 41.9 | Hurt |
| DSBCO vs MGCAMT (RetinaNet) | DSBCO · KITTI -> Cityscapes | AP of Car: 62.2 vs 62.2 | No clear effect |
| use feature-based ridge solver instead of iterative bilevel inner loops | DSBCO · Training time | time relative to Standard ST: approx M*T -> approx 1.2*T | Improved |
Closest papers by meaning
- Cycle Self-Training for Domain Adaptation (NeurIPS 2021 · similarity 0.90)
- A Free Lunch for Unsupervised Domain Adaptive Object Detection without Source Data (AAAI 2021 · similarity 0.89)
- A Robust Learning Approach to Domain Adaptive Object Detection (ICCV 2019 · similarity 0.89)
- Self-Training and Adversarial Background Regularization for Unsupervised Domain Adaptive One-Stage Object Detection (ICCV 2019 · similarity 0.88)
- Robust Object Detection via Instance-Level Temporal Cycle Confusion (ICCV 2021 · similarity 0.88)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D 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.
