Research map / Detection and segmentation / Open-set and few-shot detection
Dense Gaussian Processes for Few-Shot Segmentation
Joakim Johnander, Johan Edstedt, Michael Felsberg, Fahad Shahbaz Khan, Martin Danelljan
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 |
|---|---|---|---|
| use SE kernel instead of linear kernel (ablation) | DGPNet · PASCAL-5i 5-shot | mIoU: 61.8 -> 69.9 | Improved |
| use exponential kernel instead of linear kernel (ablation) | DGPNet · COCO-20i 5-shot | mIoU: 45.3 -> 51.8 | Improved |
| add predictive covariance (5x5 window) to decoder input (ablation) | DGPNet · PASCAL-5i and COCO-20i mean | mIoU: gain of 1.7 mIoU | Improved |
| add mask encoder to learn the GP output space (ablation) | DGPNet · PASCAL-5i and COCO-20i mean | mIoU: 1.7 improvement in isolation, or a 2.5 mIoU gain together with the covariance | Improved |
| add covariance only on 1-shot PASCAL (ablation) | DGPNet · PASCAL-5i 1-shot | mIoU: 62.1 -> 62.5 | No clear effect |
| add stride-16 GP alongside stride-32 GP (ablation) | DGPNet · PASCAL-5i 5-shot | mIoU: 72.6 -> 73.6 | Improved |
| use stride-16 only instead of stride-32 (ablation) | DGPNet · PASCAL-5i 5-shot | mIoU: 72.6 -> 69.3 | Hurt |
| increase covariance window from 5x5 to 7x7 (ablation) | DGPNet · PASCAL-5i and COCO-20i mean | mIoU: 1.7 -> 1.8 | No clear effect |
| use 1x1 covariance window instead of none (ablation) | DGPNet · PASCAL-5i 1-shot | mIoU: 62.1 -> 60.0 | Hurt |
| fine-tune image encoder instead of freezing it (ablation) | DGPNet · COCO-20i 5-shot | mIoU: 54.5 -> 56.2 | Improved |
| use GP instead of prototype-based support aggregation (ablation) | DGPNet · PASCAL-5i 5-shot | mIoU: 63.0 -> 73.5 | Improved |
| use GP instead of PFENet prior mask (ablation) | DGPNet · COCO-20i 5-shot | mIoU: 37.2 -> 56.2 | Improved |
Closest papers by meaning
- Feature Weighting and Boosting for Few-Shot Segmentation (ICCV 2019 · similarity 0.88)
- Few-Shot Segmentation Without Meta-Learning: A Good Transductive Inference Is All You Need? (CVPR 2021 · similarity 0.88)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.86)
- CGMGM: A Cross-Gaussian Mixture Generative Model for Few-Shot Semantic Segmentation (AAAI 2024 · similarity 0.86)
- A Surprisingly Simple Approach to Generalized Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.86)
- Task-Disruptive Background Suppression for Few-Shot Segmentation (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.
