Research map / Detection and segmentation / Open-set and few-shot detection
Remedying Target-Domain Astigmatism for Cross-Domain Few-Shot Object Detection
Yongwei Jiang, Yixiong Zou, Yuhua Li, Ruixuan Li
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 NCM (negative context modulation) (ablation) | GLIP Swin-T · 6 CD-FSOD datasets, 5-shot | mAP: +1.06% avg | Improved |
| add PPR (positive pattern refinement) (ablation) | GLIP Swin-T · 6 CD-FSOD datasets, 5-shot | mAP: +2.06% avg | Improved |
| add TSA (textual semantic alignment) (ablation) | GLIP Swin-T · 6 CD-FSOD datasets, 5-shot | mAP: +1.59% avg | Improved |
| add NCM, PPR and TSA together (ablation) | GLIP Swin-T · DeepFish 5-shot | mAP: 28.40 -> 33.87 | Improved |
| add NCM, PPR and TSA together (ablation) | GLIP Swin-T · UODD 5-shot | mAP: 11.40 -> 15.66 | Improved |
| apply full method to GLIP in 1-shot | GLIP Swin-T · 6 datasets 1-shot avg | mAP: 18.98 -> 23.81 | Improved |
| apply full method to GLIP in 10-shot | GLIP Swin-T · 6 datasets 10-shot avg | mAP: 33.93 -> 39.17 | Improved |
| increase background text count from 50 to 200 (ablation) | TSA module · text count study | AP: 39.34 -> 42.87 | Improved |
| increase background text count from 200 to 400 (ablation) | TSA module · text count study | AP: 42.87 -> 44.05 | No clear effect |
| increase GPU memory with more background texts (ablation) | TSA module · text count study | GPU memory: 8762 MB -> 9245 MB | Hurt |
| set TSA loss weight to 10^3 versus extremes (ablation) | TSA loss · parameter analysis | detection performance | Mixed |
| set NCM threshold to 0.75 versus lower thresholds (ablation) | NCM · parameter analysis | detection performance | Mixed |
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.
- Domain-RAG* — Domain-RAG: Retrieval-Guided Compositional Image Generation for Cross-Domain Few-Shot Object Detection (NeurIPS 2025) · better on 3 of 3 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
- Cross-Domain Few-Shot Object Detection via Enhanced Open-Set Object Detector (ECCV 2024 · similarity 0.89)
- Cross-domain Few-shot Object Detection with Multi-modal Textual Enrichment (IJCV 2026 · similarity 0.88)
- Domain-RAG: Retrieval-Guided Compositional Image Generation for Cross-Domain Few-Shot Object Detection (NeurIPS 2025 · similarity 0.88)
- SEEN-DA: SEmantic ENtropy guided Domain-aware Attention for Domain Adaptive Object Detection (CVPR 2025 · similarity 0.88)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.86)
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.
