Research map / Detection and segmentation / Open-set and few-shot detection
AMP: Adaptive Masked Proxies for Few-Shot Segmentation
Mennatullah Siam, Boris N. Oreshkin, Martin Jagersand
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 proxy adaptation (alpha > 0) (ablation) | AMP imprinting · PASCAL-5i fold 0 (1-shot) | mIoU: 13.6 -> 41.9 | Improved |
| add multi-resolution imprinting (ablation) | AMP imprinting · PASCAL-5i fold 0 (1-shot) | mIoU: 34.8 -> 41.9 | Improved |
| imprint weights with masked proxies instead of fine-tuning random weights (ablation) | AMP · PASCAL-5i fold 0 (5-shot) | mIoU: 28.7 -> 40.3 | Improved |
| add fine-tuning after imprinting (ablation) | AMP · PASCAL-5i fold 0 (5-shot) | mIoU: 40.3 -> 41.8 | Improved |
| add 2 fine-tuning iterations | AMP-2 · PASCAL-5i 5-shot (mean) | mIoU: 46.4 -> 46.9 | Improved |
| add 2 fine-tuning iterations | AMP-2 · PASCAL-5i 1-shot (FG/BG protocol) | mIoU: 61.9 -> 62.2 | Improved |
| add 2 fine-tuning iterations | AMP-2 · PASCAL-5i 5-shot (FG/BG protocol) | mIoU: 62.1 -> 63.8 | Improved |
| use Reduced-DFCN8s backbone instead of DFCN8s | AMP · PASCAL-5i 1-shot (mean) | mIoU: 42.4 -> 43.4 | Improved |
| replace OSLSM two-branch method with AMP-2 | few-shot segmentation · PASCAL-5i 1-shot (mean) | mIoU: 40.8 -> 43.4 | Improved |
| imprinted weights instead of fine-tuning in continual setting | AMP · iPASCAL | mIoU | Improved |
| self-adaptation with AMP plus CRF on 2-stream network | video object segmentation · DAVIS'16 | J mean: 77.2 -> 78.9 | Improved |
| self-adaptation with AMP | video object segmentation · FBMS test | F-measure: 79.0 -> 79.0 | No clear effect |
Closest papers by meaning
- Self-Guided and Cross-Guided Learning for Few-Shot Segmentation (CVPR 2021 · similarity 0.88)
- HM: Hybrid Masking for Few-Shot Segmentation (ECCV 2022 · similarity 0.87)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.87)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.86)
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.86)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 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.
