Research map / Detection and segmentation / Open-set and few-shot detection
SGQA: Semantic-Geometric Quality Alignment for Training-Free Few-Shot Instance Segmentation
Yuhao Qing, Liuyan Feng, Haoyuan 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 |
|---|---|---|---|
| replace decoupled scoring with geometric mean of semantic and geometric scores | NTT (DINOv2 ViT-L + SAM2 Hiera-L) · COCO 30-shot | Bbox nAP: 36.8 -> 39.4 | Improved |
| use arithmetic mean instead of geometric mean for fusion (ablation) | SGQA · COCO 30-shot | AP: +1.1 AP vs +2.6 AP | Hurt |
| add temperature scaling calibration of scores (ablation) | NTT · COCO 30-shot | AP: +0.5 AP | Improved |
| use harmonic-style penalties (p<0) in power mean (ablation) | SGQA · COCO 30-shot | AP | Hurt |
| vary alpha weighting away from 0.5 toward semantic-only or geometric-only (ablation) | SGQA · COCO 30-shot | AP | Hurt |
| apply unified score only to NMS ranking (ablation) | SGQA · COCO 30-shot | AP: +1.7 AP | Improved |
| apply unified score only as detection confidence (ablation) | SGQA · COCO 30-shot | AP: +1.3 AP | Improved |
| add SGQA to different encoders (CLIP, DINOv2, DINOv3) (ablation) | frozen composition with eight encoders · COCO 30-shot | Bbox AP: +1.6 to +2.6 AP | Improved |
| add SGQA to a smaller mask generator (SAM2-Tiny) (ablation) | NTT · COCO 30-shot | Bbox AP: 35.2 with SGQA vs 36.8 for SAM2-Large baseline | Improved |
| reduce number of shots from 30 to 1 (ablation) | SGQA gain over NTT · COCO few-shot | Bbox nAP gain: +2.6 -> +3.3 | Improved |
| increase shots from 10 to 30 (ablation) | SGQA · COCO-FSOD | Bbox nAP: 39.3 -> 39.4 | No clear effect |
| add SGQA on PASCAL VOC | NTT · PASCAL VOC novel AP50 | AP50: 71.2 -> 77.4 | Improved |
Closest papers by meaning
- Mitigating Pose–Scale Discrepancy Bias and Reforming Multi-Support Reasoning for Few-Shot Semantic Segmentation (ECCV 2026 · similarity 0.89)
- SGS-3D: High-Fidelity 3D Instance Segmentation via Reliable Semantic Mask Splitting and Growing (AAAI 2026 · similarity 0.87)
- Geometry-Guided Semantic Reconstruction for 3D Instance Segmentation (NeurIPS 2026 · similarity 0.87)
- GeoGuide: Hierarchical Geometric Guidance for Open-Vocabulary 3D Semantic Segmentation (CVPR 2026 · similarity 0.86)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.86)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · 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.
