Research map / Detection and segmentation / Weakly supervised detection and segmentation
Normalized Cut Loss for Weakly-Supervised CNN Segmentation
Read the paperCheck 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 normalized cut loss to partial cross entropy training (ablation) | DeepLab-MSc-largeFOV · PASCAL VOC 2012 val | mIOU (%): 56.0 -> 60.5 | Helped |
| Add normalized cut loss to partial cross entropy training (ablation) | DeepLab-VGG16 · PASCAL VOC 2012 val | mIOU (%): 60.4 -> 62.4 | Helped |
| Add normalized cut loss to partial cross entropy training (ablation) | DeepLab-ResNet-101 · PASCAL VOC 2012 val | mIOU (%): 69.5 -> 72.8 | Helped |
| Add normalized cut loss to pCE with multi-scale branches (ablation) | DeepLab-MSc-largeFOV · PASCAL VOC 2012 val | mIOU (%): 56.1 -> 59.7 | Helped |
| Add the non-existent-label penalty (ablation) | pCE + MSC + NC + CRF · PASCAL VOC 2012 val | mIOU (%): 60.5 -> 65.1 | Helped |
| Add multi-scale branches to pCE training (ablation) | DeepLab-MSc-largeFOV · PASCAL VOC 2012 val | mIOU (%): 55.8 -> 56.1 | Helped |
| Add CRF post-processing to pCE + MSC + NC (ablation) | DeepLab-MSc-largeFOV · PASCAL VOC 2012 val | mIOU (%): 59.7 -> 60.5 | Helped |
| Train using partial cross entropy on scribbles instead of GrabCut proposals | DeepLab-largeFOV · PASCAL VOC 2012 val | mIOU (%): 54.7 -> 55.8 | Helped |
Closest papers by meaning
- Beyond Gradient Descent for Regularized Segmentation Losses (CVPR 2019 · similarity 0.89)
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.88)
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.88)
- CIAN: Cross-Image Affinity Net for Weakly Supervised Semantic Segmentation (AAAI 2020 · similarity 0.88)
- Box-Driven Class-Wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation (CVPR 2019 · similarity 0.88)
- Soft Self-labeling and Potts Relaxations for Weakly-supervised Segmentation (CVPR 2025 · 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.