Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Singular Value Fine-tuning: Few-shot Segmentation requires Few-parameters Fine-tuning
Yanpeng Sun, Qiang Chen, Xiangyu He, Jian Wang, Haocheng Feng, Junyu Han, Errui Ding, Jian Cheng, Zechao Li, 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 |
|---|---|---|---|
| fine-tune singular values S only (SVF) on frozen backbone | baseline (ResNet-50) · Pascal-5i 1-shot | mIoU: 65.07 -> 67.14 | Improved |
| add SVF | BAM (ResNet-50) · Pascal-5i 1-shot | mIoU: 67.81 -> 68.95 | Improved |
| add SVF | BAM (ResNet-50) · COCO-20i 5-shot | mIoU: 51.16 -> 53.87 | Improved |
| add SVF | BAM (VGG-16) · COCO-20i 5-shot | mIoU: 49.34 -> 49.07 | Hurt |
| add SVF | PFENet (ResNet-50) · Pascal-5i 5-shot | mIoU: 71.20 -> 71.82 | Improved |
| fully fine-tune the backbone instead of freezing (ablation) | baseline · Pascal-5i 1-shot | mIoU: 65.07 -> 60.90 | Hurt |
| fine-tune U or V subspace instead of S (ablation) | SVF · Pascal-5i 1-shot | mIoU: 67.14 -> 61.09 (U), 60.88 (V) | Hurt |
| fine-tune S only rather than U,S,V (ablation) | SVF · Pascal-5i 1-shot | mIoU: 61.24 -> 67.14 | Improved |
| also fine-tune BN with SVF (ablation) | SVF · Pascal-5i 1-shot | mIoU: 67.14 -> 64.20 | Hurt |
| replace SVF with adapter (ablation) | baseline · Pascal-5i 1-shot | mIoU: 67.14 -> 20.71 | Hurt |
| replace SVF with BN bias tuning (ablation) | baseline · Pascal-5i 1-shot | mIoU: 67.14 -> 62.93 | Hurt |
| apply SVF to layers 3,4 vs 2,3,4 vs 1-4 (ablation) | baseline · Pascal-5i 1-shot | mIoU: 67.20 / 67.14 / 67.12 | No clear effect |
Closest papers by meaning
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.88)
- Strong Baselines for Parameter-Efficient Few-Shot Fine-Tuning (AAAI 2024 · similarity 0.87)
- A Surprisingly Simple Approach to Generalized Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.86)
- The Devil is in Low-Level Features for Cross-Domain Few-Shot Segmentation (CVPR 2025 · similarity 0.86)
- SAM-PARSER: Fine-Tuning SAM Efficiently by Parameter Space Reconstruction (AAAI 2024 · similarity 0.86)
- Tuning Stable Rank Shrinkage: Aiming at the Overlooked Structural Risk in Fine-tuning (CVPR 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.
