Research map / Detection and segmentation / Open-set and few-shot detection
Balanced and Hierarchical Relation Learning for One-Shot 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 |
|---|---|---|---|
| Replace a single relation network with the three-level IHR module (ablation) | BHRL · COCO split-2 unseen classes | AP50: 25.0 -> 27.9 | Helped |
| Replace a single relation network with the three-level IHR module (ablation) | BHRL · PASCAL VOC unseen classes | AP50: 67.5 -> 71.7 | Helped |
| Use RPL instead of CEL (ablation) | BHRL with a relation network · COCO split-2 unseen classes | AP50: 25.0 -> 27.5 | Helped |
| Use RPL instead of CEL (ablation) | BHRL with a relation network · PASCAL VOC unseen classes | AP50: 67.5 -> 70.9 | Helped |
| Fuse contrastive-, salient-, and attention-level relation features (ablation) | IHR · COCO split-2 unseen classes | AP50: 26.4 / 26.6 / 26.7 -> 27.9 | Helped |
| Add the target feature to the three fused relation levels (ablation) | IHR · COCO split-2 unseen classes | AP50: 27.6 -> 27.9 | Helped |
| Use score addition rather than feature fusion for the three relation levels (ablation) | IHR · COCO split-2 unseen classes | AP50: 27.2 vs 27.6 | Hurt |
| Set RPL α to 1/4 (ablation) | BHRL · COCO split-2 unseen classes | AP50: 27.9 -> 29.0 | Helped |
| Replace CEL with Focal Loss at classification-task weight 1 (ablation) | BHRL · COCO split-2 unseen classes | AP50: 27.9 -> 25.8 | Hurt |
| Use two query patches instead of one at inference (ablation) | BHRL · COCO split-1 seen classes | AP50: 56.0 -> 57.8 | Helped |
Closest papers by meaning
- Adaptive Image Transformer for One-Shot Object Detection (CVPR 2021 · similarity 0.87)
- UniT: Unified Knowledge Transfer for Any-Shot Object Detection and Segmentation (CVPR 2021 · similarity 0.86)
- Semantic Relation Reasoning for Shot-Stable Few-Shot Object Detection (CVPR 2021 · similarity 0.86)
- One-Shot Object Detection with Co-Attention and Co-Excitation (NeurIPS 2019 · similarity 0.86)
- Exploring Base-Class Suppression with Prior Guidance for Bias-Free One-Shot Object Detection (AAAI 2024 · similarity 0.85)
- CQ-DINO: Mitigating Gradient Dilution via Category Queries for Vast Vocabulary Object Detection (NeurIPS 2025 · similarity 0.85)
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.