Research map / Detection and segmentation / Open-set and few-shot detection
Rethinking Prototype-based Similarity Learning for Few-Shot Object Detection
KunHo Heo, Seungjae Kim, Wongyu Lee, SuYeon Kim, MyeongAh Cho
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 SHARe hierarchical autoregressive regression (ablation) | prototype baseline · COCO 30-shot | nAP75: 36.8 -> 47.4 | Improved |
| add TSMa text-anchored semantic mask (ablation) | prototype baseline · COCO 30-shot | nAP50: 53.6 -> 63.5 | Improved |
| add both TSMa and SHARe (ablation) | prototype baseline · COCO 30-shot | nAP: 34.3 -> 47.4 | Improved |
| add TF-IDF scoring in TSMa (ablation) | TSMa · COCO 30-shot | nAP: 45.8 -> 47.4 | Improved |
| inject ViT features in reverse order instead of none (ablation) | SHARe · COCO 30-shot | nAP: 43.7 -> 45.9 | Improved |
| use reverse injection instead of forward (ablation) | SHARe · COCO 30-shot | nAP75: 48.2 -> 51.5 | Improved |
| learnable weighted sum instead of uniform average aggregation (ablation) | SHARe · COCO 30-shot | nAP: 46.6 -> 47.4 | Improved |
| learnable weighted sum instead of max pooling (ablation) | SHARe · COCO 30-shot | nAP: 46.2 -> 47.4 | Improved |
| use 5 stages instead of 3 (ablation) | SHARe · COCO 30-shot | nAP: 46.6 -> 47.4 | Improved |
| use 5 stages instead of 3 (ablation) | SHARe · COCO 30-shot | nAP50: 65.9 -> 65.6 | Hurt |
| add TSMa in one-shot (ablation) | prototype baseline · COCO one-shot split-4 | nAP50: 32.5 -> 37.9 | Improved |
| scale backbone from ViT-S/14 to ViT-L/14 (ablation) | Ours · COCO 30-shot | nAP: 35.5 -> 47.4 | Improved |
Closest papers by meaning
- Stabilizing Few-Shot Object Detection with Language-Conditioned Probabilistic Prototypes (NeurIPS 2026 · similarity 0.91)
- StyleProto: Style-Augmented Prototype Learning for Cross-Domain Few-Shot Object Detection (AAAI 2026 · similarity 0.88)
- Cross-domain Few-shot Object Detection with Multi-modal Textual Enrichment (IJCV 2026 · similarity 0.87)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.86)
- Text Augmented Correlation Transformer For Few-shot Classification & Segmentation (CVPR 2025 · 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.
