Research map / Label-efficient and robust learning / Domain adaptation and generalization
Target-Relevant Knowledge Preservation for Multi-Source Domain Adaptive Object Detection
Check 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 |
|---|---|---|---|
| Add AMSD to the UBT baseline | UBT · BDD100K dawn/dusk (daytime+night sources) | mAP: 33.1 -> 38.0 | Helped |
| Add HTRM to the UBT baseline | UBT · BDD100K dawn/dusk (daytime+night sources) | mAP: 33.1 -> 35.5 | Helped |
| Add AMSD classification and regression disentanglement (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 34.4 -> 38.0 | Helped |
| Use separate detectors per source instead of a shared feature extractor (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 33.1 -> 24.7 | Hurt |
| Add multi-head detection to shared features (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 33.1 -> 34.4 | Helped |
| Apply AMSD to classification head only (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 34.4 -> 36.5 | Helped |
| Apply AMSD to regression head only (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 34.4 -> 35.9 | Helped |
| Set AMSD μ to 0.05 instead of 0.01 (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 38.0 -> 36.8 | Hurt |
| Set AMSD μ to 0.002 instead of 0.01 (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 38.0 -> 37.0 | Hurt |
| Set AMSD λ to 1.0 instead of 0.2 (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 38.0 -> 37.2 | Hurt |
| Set AMSD λ to 0.04 instead of 0.2 (ablation) | TRKP · BDD100K dawn/dusk (daytime+night sources) | mAP: 38.0 -> 36.7 | Hurt |
| Mine source-target relevance at image level rather than instance level (ablation) | HTRM · BDD100K dawn/dusk (daytime+night sources) | mAP: 35.5 -> 32.6 | Hurt |
Closest papers by meaning
- DA-Ada: Learning Domain-Aware Adapter for Domain Adaptive Object Detection (NeurIPS 2024 · similarity 0.90)
- SEEN-DA: SEmantic ENtropy guided Domain-aware Attention for Domain Adaptive Object Detection (CVPR 2025 · similarity 0.90)
- Large Self-Supervised Models Bridge the Gap in Domain Adaptive Object Detection (CVPR 2025 · similarity 0.89)
- Triple Feature Disentanglement for One-Stage Adaptive Object Detection (AAAI 2024 · similarity 0.88)
- Universal Domain Adaptive Object Detection via Dual Probabilistic Alignment (AAAI 2025 · similarity 0.88)
- DSD-DA: Distillation-based Source Debiasing for Domain Adaptive Object Detection (ICML 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.