Research map / Detection and segmentation / Open-set and few-shot detection
HM: Hybrid Masking for Few-Shot Segmentation
Seonghyeon Moon, Samuel S. Sohn, Honglu Zhou, Sejong Yoon, Vladimir Pavlovic, Muhammad Haris Khan, Mubbasir Kapadia
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 feature masking with hybrid masking | HSNet ResNet50 · COCO-20i 1-shot | mIoU: 39.2 -> 44.3 | Improved |
| replace feature masking with hybrid masking | HSNet ResNet101 · COCO-20i 1-shot | mIoU: 41.2 -> 46.5 | Improved |
| replace feature masking with hybrid masking | VAT ResNet50 · COCO-20i 1-shot | mIoU: 41.3 -> 43.2 | Improved |
| replace feature masking with hybrid masking | ASNet ResNet50 · COCO-20i 5-shot | mIoU: 49.5 -> 48.4 | Hurt |
| replace feature masking with hybrid masking | HSNet* ResNet50 · PASCAL-5i 5-shot | mIoU: 69.6 -> 67.1 | Hurt |
| replace feature masking with hybrid masking | VAT ResNet50 · PASCAL-5i 1-shot | mIoU: 65.3 -> 65.8 | Improved |
| replace feature masking with hybrid masking | VAT ResNet50 · FSS-1000 1-shot | mIoU: 89.5 -> 89.4 | Hurt |
| replace feature masking with hybrid masking | HSNet ResNet50 · FSS-1000 1-shot | mIoU: 85.5 -> 87.1 | Improved |
| replace feature masking with hybrid masking | HSNet ResNet50 · COCO-20i to PASCAL-5i domain shift 1-shot | mIoU: 61.6 -> 65.2 | Improved |
| add FM features to IM through hybrid masking (versus input masking alone) (ablation) | HSNet ResNet50 · COCO-20i 1-shot | mIoU: 43.5 -> 44.3 | Improved |
| fill inactive FM pixels with IM features (versus simple addition) (ablation) | HSNet-HM ResNet50 · COCO-20i 1-shot | mIoU: 42.5 -> 44.3 | Improved |
| use FM as the primary map and IM as supplement (versus reverse) (ablation) | HSNet-HM ResNet50 · COCO-20i 1-shot | mIoU: 42.1 -> 44.3 | Improved |
Closest papers by meaning
- Mask Matching Transformer for Few-Shot Segmentation (NeurIPS 2022 · similarity 0.90)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.89)
- Lightweight Frequency Masker for Cross-Domain Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.88)
- Foreground-Covering Prototype Generation and Matching for SAM-Aided Few-Shot Segmentation (AAAI 2025 · similarity 0.88)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 2024 · similarity 0.87)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 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.
