Research map / Detection and segmentation / Open-set and few-shot detection
FSS-1000: A 1000-Class Dataset for Few-Shot Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the adapted relation network instead of OSLSM or Guided Network | Few-shot segmentation trained and tested on FSS-1000 · FSS-1000 | MeanIoU: 80.12% vs 74.27% | Helped |
| Increase the support set from 1 image to 5 (ablation) | Ours few-shot segmentation · FSS-1000 | MeanIoU: 73.47% -> 80.12% | Helped |
| Train on FSS-1000 rather than PASCAL-5^i | Ours few-shot segmentation evaluated on PASCAL-5^i · PASCAL-5^i | MeanIoU: 46.8% -> 58.6% | Helped |
| Pre-train on FSS-1000, then fine-tune on fsCOCO | VGG-16 few-shot segmentation · FSS test set | MeanIoU: 71.34% -> 82.66% | Helped |
| Pre-train on FSS-1000, then fine-tune on fsCOCO | VGG-16 few-shot segmentation · fsCOCO test set | MeanIoU: 42.11% -> 50.56% | Helped |
| Remove ImageNet pre-training and train exclusively on FSS-1000 (ablation) | VGG-16 few-shot segmentation · FSS test set | MeanIoU: 79.30% -> 80.12% | Helped |
| Remove ImageNet pre-training and train exclusively on FSS-1000 (ablation) | VGG-16 few-shot segmentation · fsCOCO test set | MeanIoU: 47.99% -> 48.31% | Helped |
| Replace BCE loss with MSE loss while retaining VGG-16 (ablation) | Few-shot segmentation trained on FSS-1000 · FSS-1000 | MeanIoU: 80.12% -> 79.66% | Hurt |
| Replace VGG-16 with ResNet-101 while retaining BCE loss (ablation) | Few-shot segmentation trained on FSS-1000 · FSS-1000 | MeanIoU: 80.12% -> 79.43% | Hurt |
| Replace VGG-16 with InceptionV3 while retaining BCE loss (ablation) | Few-shot segmentation trained on FSS-1000 · FSS-1000 | MeanIoU: 80.12% -> 79.02% | Hurt |
| Use MSE instead of BCE across the tested backbones (ablation) | Few-shot segmentation trained on FSS-1000 · FSS-1000 | MeanIoU: 79.43% -> 79.12%; 79.02% -> 79.22% | Mixed |
| Increase support examples beyond 5 (ablation) | Few-shot segmentation model · FSS-1000 | MeanIoU | No clear effect |
Closest papers by meaning
- Adaptive FSS: A Novel Few-Shot Segmentation Framework via Prototype Enhancement (AAAI 2024 · similarity 0.91)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.89)
- Text Augmented Correlation Transformer For Few-shot Classification & Segmentation (CVPR 2025 · similarity 0.88)
- Unlocking the Power of SAM 2 for Few-Shot Segmentation (ICML 2025 · similarity 0.88)
- A Surprisingly Simple Approach to Generalized Few-Shot Semantic Segmentation (NeurIPS 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.