Research map / Detection and segmentation / Open-set and few-shot detection
Single-Domain Generalized Object Detection in Urban Scene via Cyclic-Disentangled Self-Distillation
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 |
|---|---|---|---|
| apply cyclic-disentangled self-distillation rather than Faster R-CNN | Faster R-CNN · Daytime-sunny | mAP: 51.1 -> 56.1 | Helped |
| apply cyclic-disentangled self-distillation rather than Faster R-CNN | Faster R-CNN · Night-sunny | mAP: 33.5 -> 36.6 | Helped |
| apply cyclic-disentangled self-distillation rather than Faster R-CNN | Faster R-CNN · Dusk-rainy | mAP: 26.6 -> 28.2 | Helped |
| apply cyclic-disentangled self-distillation rather than Faster R-CNN | Faster R-CNN · Night-rainy | mAP: 14.5 -> 16.6 | Helped |
| apply cyclic-disentangled self-distillation rather than Faster R-CNN | Faster R-CNN · Daytime-foggy | mAP: 31.9 -> 33.5 | Helped |
| add cyclic disentanglement to forward-only disentanglement (ablation) | CDSD · Daytime-sunny and Night-sunny | mAP: 48.7% -> 50.4%; 30.4% -> 31.3% | Helped |
| add self-distillation to cyclic disentanglement (ablation) | CDSD · Daytime-sunny and Night-sunny | mAP: 52.3% -> 56.1%; 34.2% -> 36.6% | Helped |
| use self-distillation without domain-invariant teacher features (ablation) | CDSD · Daytime-sunny and Night-sunny | mAP: 51.0% vs 56.1%; 33.3% vs 36.6% | Hurt |
| remove self-distillation from cyclic disentanglement (ablation) | CDSD · Daytime-sunny and Night-sunny | mAP: 56.1% -> 52.3%; 36.6% -> 34.2% | Hurt |
| use only the forward step for disentanglement (ablation) | CDSD · Daytime-sunny and Night-sunny | mAP: 48.7% vs 56.1%; 30.4% vs 36.6% | Hurt |
| use only the global-level contrastive loss or only the instance-level contrastive loss (ablation) | Cyclic-disentangled self-distillation · Night-sunny | mAP: 34.8%; 35.2% (full method: 36.6%) | Hurt |
| use only the feature-level distillation loss or only the classification-level distillation loss (ablation) | Cyclic-disentangled self-distillation · Night-sunny | mAP: 35.0%; 35.4% | Hurt |
Closest papers by meaning
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.90)
- Towards Single-Source Domain Generalized Object Detection via Causal Visual Prompts (NeurIPS 2025 · similarity 0.90)
- Triple Feature Disentanglement for One-Stage Adaptive Object Detection (AAAI 2024 · similarity 0.90)
- Object-Aware Domain Generalization for Object Detection (AAAI 2024 · similarity 0.89)
- G-NAS: Generalizable Neural Architecture Search for Single Domain Generalization Object Detection (AAAI 2024 · similarity 0.89)
- PhysAug: A Physical-guided and Frequency-based Data Augmentation for Single-Domain Generalized 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.