Research map / Detection and segmentation / Open-set and few-shot detection
CANet: Class-Agnostic Segmentation Networks With Iterative Refinement and Attentive Few-Shot Learning
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 CANet instead of OSLSM | Few-shot segmentation on PASCAL-5^i · PASCAL-5^i, 1-shot | meanIoU: 40.8 -> 55.4 | Helped |
| Use CANet instead of OSLSM | Few-shot segmentation on PASCAL-5^i · PASCAL-5^i, 5-shot | meanIoU: 43.9 -> 57.1 | Helped |
| Use block2 and block3 features together rather than block3 alone (ablation) | CANet feature comparison on PASCAL-5^i · PASCAL-5^i | meanIoU: 50.8 -> 51.2 | Helped |
| Add block4 to the block2 + block3 feature combination (ablation) | CANet feature comparison on PASCAL-5^i · PASCAL-5^i | meanIoU: 51.2 -> 49.5 | Hurt |
| Replace ResNet-50 with VGG16 features (ablation) | CANet on PASCAL-5^i · PASCAL-5^i | meanIoU: 55.4 -> 54.3 | Hurt |
| Add iterative optimization to the initial prediction (ablation) | CANet on PASCAL-5^i · PASCAL-5^i | meanIoU: 51.2 -> 54.0 | Helped |
| Replace learned iterative optimization with DenseCRF post-processing (ablation) | CANet initial prediction on PASCAL-5^i · PASCAL-5^i | meanIoU: 51.2 -> 51.9 | No clear effect |
| Fuse support features by averaging instead of using a single-shot baseline (ablation) | CANet 5-shot segmentation on PASCAL-5^i · PASCAL-5^i, 5-shot | meanIoU: 54.0 -> 55.0 | Helped |
| Fuse masks by averaging instead of using a single-shot baseline (ablation) | CANet 5-shot segmentation on PASCAL-5^i · PASCAL-5^i, 5-shot | meanIoU: 54.0 -> 54.5 | Helped |
| Fuse masks with logical OR instead of using a single-shot baseline (ablation) | CANet 5-shot segmentation on PASCAL-5^i · PASCAL-5^i, 5-shot | meanIoU: 54.0 -> 53.4 | Hurt |
| Use learned attention to fuse multiple support examples (ablation) | CANet 5-shot segmentation on PASCAL-5^i · PASCAL-5^i, 5-shot | meanIoU: 54.0 -> 55.8 | Helped |
| Use bounding-box rather than pixel-wise support annotations (ablation) | CANet on PASCAL-5^i · PASCAL-5^i, support annotation | meanIoU: 54.0 -> 52.0 | Hurt |
Closest papers by meaning
- CRNet: Cross-Reference Networks for Few-Shot Segmentation (CVPR 2020 · similarity 0.87)
- Few-Shot Segmentation Without Meta-Learning: A Good Transductive Inference Is All You Need? (CVPR 2021 · similarity 0.86)
- A Surprisingly Simple Approach to Generalized Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.86)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.85)
- Relevant Intrinsic Feature Enhancement Network for Few-Shot Semantic Segmentation (AAAI 2024 · similarity 0.85)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 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.