Research map / Detection and segmentation / CNN features
Sparse Non-Local CRF
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 |
|---|---|---|---|
| Add sparse non-local connections with distance weights to GrabCut | GrabCut segmentation · GrabCut | Fβ: .897 -> .928 | Helped |
| Add sparse non-local connections with Gaussian weights | GrabCut segmentation · MSRA1K | Fβ: .956 -> .966 | Helped |
| Add sparse non-local connections with Gaussian weights | GrabCut segmentation · ECSSD | Fβ: .868 -> .892 | Helped |
| Replace sparse CRF with sparse non-local CRF using Gaussian weights | Weakly supervised salient object segmentation · MSRAB (MSRA training) | Fβ: .885 -> .902 | Helped |
| Replace sparse CRF with sparse non-local CRF using distance weights | Weakly supervised salient object segmentation · ECSSD (MSRA training) | mae: .056 -> .056 | Helped |
| Replace sparse CRF with sparse non-local CRF using distance weights | Weakly supervised salient object segmentation · DUTO (MSRA training) | Fβ: .753 -> .771 | Helped |
| Use distance weights rather than Gaussian weights for non-local edges | GrabCut segmentation · GrabCut | Fβ: .919 -> .928 | Helped |
| Use distance weights rather than Gaussian weights for non-local edges | Weakly supervised salient object segmentation · DUTO (DUTS training) | maxFβ: .785 -> .793 | Helped |
| Use distance weights rather than Gaussian weights for non-local edges | Weakly supervised salient object segmentation · ECSSD (DUTS training) | maxFβ: .910 -> .909 | No clear effect |
| Increase the number of sampled non-local edges per pixel from 2 to 16 (ablation) | GrabCut with distance weights · GrabCut dataset | Fβ: .926 (.0024) -> .928 (.0007) | No clear effect |
| Use two color quantizations instead of one (ablation) | GrabCut with distance weights · GrabCut dataset | Fβ: .926 (.0024) -> .928 (.0012) | Helped |
| Increase sampled edge count from 2 to 32 with Gaussian weights (ablation) | GrabCut · GrabCut dataset | Fβ: .920 (.0022) -> .919 (.0026) | No clear effect |
Closest papers by meaning
- Efficient Linear Programming for Dense CRFs (CVPR 2017 · similarity 0.86)
- Regularized Frank-Wolfe for Dense CRFs: Generalizing Mean Field and Beyond (NeurIPS 2021 · similarity 0.86)
- Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation (CVPR 2018 · similarity 0.86)
- Gaussian Conditional Random Field Network for Semantic Segmentation (CVPR 2016 · similarity 0.85)
- Multi-Scale Continuous CRFs as Sequential Deep Networks for Monocular Depth Estimation (CVPR 2017 · similarity 0.85)
- End-To-End Training of Hybrid CNN-CRF Models for Stereo (CVPR 2017 · similarity 0.84)
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.