Research map / Detection and segmentation / Open-set and few-shot detection
Focus on Query: Adversarial Mining Transformer for Few-Shot Segmentation
Yuan Wang, Naisong Luo, Tianzhu Zhang
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 object mining transformer G to the query-feature baseline (ablation) | AMFormer · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 65.0 -> 66.5 | Improved |
| Use multi-scale rather than single-scale object mining (ablation) | AMFormer · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 66.5 -> 69.5 | Improved |
| Add detail mining transformer D without diversity loss (ablation) | AMFormer with multi-scale G · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 69.5 -> 69.7 | Improved |
| Add diversity loss to detail mining transformer D (ablation) | AMFormer with multi-scale G · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 69.5 -> 70.7 | Improved |
| Increase local proxy count from 4 to 10 (ablation) | AMFormer · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 70.7 -> 71.3 | Improved |
| Increase local proxy count beyond 10 to 12 or 14 (ablation) | AMFormer · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 71.3 -> 71.2 | No clear effect |
| Replace mask support labels with bounding-box labels (ablation) | AMFormer · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 71.3 -> 70.5 | Hurt |
| Replace mask support labels with scribble labels (ablation) | AMFormer · Pascal-5^i first split, ResNet-101, 1-shot | mIoU: 71.3 -> 70.0 | Hurt |
| Add ensemble strategy (ablation) | AMFormer · Pascal-5^i, ResNet-101, 1-shot | mIoU: 69.3 -> 70.7 | Improved |
| Erode support foreground features across the tested ratios (ablation) | PFENet and HDMNet · Pascal-5^i first split, ResNet-50, 1-shot | mIoU: PFENet: 60.6, 61.0, 61.5, 61.2, 61.0, 61.2; HDMNet: 69.3, 70.0, 70.5, 70.4, 69.9, 70.2 | No clear effect |
| Use AMFormer instead of the strongest reported baseline | Few-shot segmentation · PASCAL-5^i 1-shot, Res-101 | mIoU: 67.9 -> 70.7 | Improved |
| Use AMFormer instead of the strongest reported baseline | Few-shot segmentation · PASCAL-5^i 5-shot, Res-50 | mIoU: 71.8 -> 73.2 | Improved |
Closest papers by meaning
- Bridge the Intra-Class Gap: K-Shot Multi-Scale Intermediate Prototype Mining Transformer for Few-Shot Semantic Segmentation (TPAMI 2025 · similarity 0.89)
- Mining Latent Classes for Few-Shot Segmentation (ICCV 2021 · similarity 0.89)
- Intermediate Prototype Mining Transformer for Few-Shot Semantic Segmentation (NeurIPS 2022 · similarity 0.88)
- Cross-Domain Few-Shot Segmentation via Iterative Support-Query Correspondence Mining (CVPR 2024 · similarity 0.86)
- Foreground-Covering Prototype Generation and Matching for SAM-Aided Few-Shot Segmentation (AAAI 2025 · similarity 0.86)
- Adaptive FSS: A Novel Few-Shot Segmentation Framework via Prototype Enhancement (AAAI 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.
