Research map / Detection and segmentation / Open-set and few-shot detection
s-Adaptive Decoupled Prototype for Few-Shot Object Detection
Jinhao Du, Shan Zhang, Qiang Chen, Haifeng Le, Yanpeng Sun, Yao Ni, Jian Wang, Bin He, Jingdong Wang
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 |
|---|---|---|---|
| replace K-average prototype with sigma-adaptive weighting (angle plus magnitude dispersion) (ablation) | FSOD baseline · FSOD dataset (5-shot, novel) | mAP: 23.1 -> 28.1 | Improved |
| use cosine similarity only or sigma only instead of combining (ablation) | sigma-ADP · FSOD dataset (5-shot, novel) | mAP: 24.2 / 24.8 vs 28.1 | Mixed |
| refine prototype twice instead of once (ablation) | sigma-ADP · FSOD dataset (5-shot, novel) | mAP: 28.1 -> 28.0 | No clear effect |
| decouple the prototype into spatial and channel task-specific ones (ablation) | sigma-ADP · FSOD dataset (5-shot, novel) | mAP: 28.0 -> 29.9 | Improved |
| train with sigma-AP instead of K-avg, test with simple K-avg (ablation) | FSOD baseline · FSOD dataset (5-shot, novel) | mAP: 23.1 -> 28.1 | Improved |
| also apply sigma-AP at test time (ablation) | sigma-ADP · FSOD dataset (5-shot, novel) | mAP: 28.1 -> 28.7 | No clear effect |
| increase support crops at inference beyond 3 (ablation) | Ours+FSOD · FSOD dataset (5-shot, novel) | mAP: 30.3 -> 28.8 | Hurt |
| swap ResNet-101 for Swin-B encoder with sigma-ADP | FSOD baseline · PASCAL VOC 5-shot | AP50: 57.1 -> 63.1 | Improved |
| apply sigma-ADP on top of DCNet | DCNet · MS COCO (10-shot, novel) | mAP: 12.8 -> 20.3 | Improved |
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.
- DeFRCN — DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection (ICCV 2021) · 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
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.90)
- Meta-Learning to Detect Rare Objects (ICCV 2019 · similarity 0.89)
- Stabilizing Few-Shot Object Detection with Language-Conditioned Probabilistic Prototypes (NeurIPS 2026 · similarity 0.88)
- A Closer Look at Cross-Domain Few-Shot Object Detection: Fine-Tuning Matters and Parallel Decoder Helps (CVPR 2026 · similarity 0.88)
- Meta-DETR: Image-Level Few-Shot Detection with Inter-Class Correlation Exploitation (TPAMI 2022 · similarity 0.88)
- SNIDA: Unlocking Few-Shot Object Detection with Non-linear Semantic Decoupling Augmentation (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.
