Research map / Detection and segmentation / Open-set and few-shot detection
MSI: Maximize Support-Set Information for Few-Shot Segmentation
Seonghyeon Moon, Samuel S. Sohn, Honglu Zhou, Sejong Yoon, Vladimir Pavlovic, Muhammad Haris Khan, Mubbasir Kapadia
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 MSI (masked support image plus full support image correlation maps) to VAT | VAT ResNet50 · PASCAL-5i 1-shot | mIoU: 65.3 -> 68.3 | Improved |
| add MSI to VAT | VAT ResNet50 · COCO-20i 1-shot | mIoU: 41.3 -> 46.8 | Improved |
| add MSI to HSNet | HSNet ResNet50 · PASCAL-5i 5-shot | mIoU: 69.5 -> 67.9 | Hurt |
| add MSI to ASNet | ASNet ResNet50 · PASCAL-5i 5-shot | mIoU: 70.8 -> 69.3 | Hurt |
| add MSI to ASNet | ASNet ResNet50 · COCO-20i 5-shot | mIoU: 47.9 -> 47.8 | No clear effect |
| add MSI to VAT, evaluate on FSS-1000 | VAT ResNet101 · FSS-1000 1-shot | mIoU: 90.0 -> 90.6 | Improved |
| add MSI to VAT, train on COCO-20i and test on PASCAL-5i | VAT ResNet50 · COCO-20i to PASCAL-5i 1-shot | mIoU: 64.5 -> 67.8 | Improved |
| add MSI to VAT on small support masks (<5% of image) | VAT ResNet50 · PASCAL-5i small masks | mIoU: 47.5 -> 51.4 | Improved |
| use full support image features (SIF) together with STF rather than STF alone (ablation) | VAT ResNet50 · PASCAL-5i 1-shot | mIoU: 65.2 -> 68.3 | Improved |
| keep feature masking in addition to STF and SIF (ablation) | VAT + MSI ResNet50 · PASCAL-5i 1-shot | mIoU: 68.3 -> 67.1 | Hurt |
| concatenate the two correlation maps instead of feature addition, correlation addition or attention fusion (ablation) | VAT + MSI ResNet50 · PASCAL-5i 1-shot | mIoU: 61.2 / 65.6 / 67.4 -> 68.3 | Improved |
| keep person class as support-image background (do not remove its mask) (ablation) | VAT + MSI ResNet50 · PASCAL-5i 1-shot | mIoU: 66.3 -> 68.3 | Improved |
Closest papers by meaning
- Adaptive FSS: A Novel Few-Shot Segmentation Framework via Prototype Enhancement (AAAI 2024 · similarity 0.89)
- Unlocking the Power of SAM 2 for Few-Shot Segmentation (ICML 2025 · similarity 0.87)
- Foreground-Covering Prototype Generation and Matching for SAM-Aided Few-Shot Segmentation (AAAI 2025 · similarity 0.86)
- Text Augmented Correlation Transformer For Few-shot Classification & Segmentation (CVPR 2025 · similarity 0.86)
- Lightweight Frequency Masker for Cross-Domain Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.86)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 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.
