Research map / Detection and segmentation / Weakly supervised detection and segmentation
Exploiting Saliency for Object Segmentation From Image Level Labels
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 |
|---|---|---|---|
| Combine class-specific seeds with class-agnostic saliency by propagating seed labels through foreground components (ablation) | G₂ guided segmentation · PASCAL VOC 2012 val | mIoU: 46.2→51.2 | Helped |
| Add an image classifier to assign labels to saliency components (ablation) | G₀ saliency-only guide · PASCAL VOC 2012 val | mIoU: 45.8→46.2 | No clear effect |
| Add saliency guidance to seed-only labels (ablation) | Guided segmentation · PASCAL VOC 2012 val, guide strategies without CRF post-processing | mIoU: 38.7→51.2 | Helped |
| Replace predicted saliency with ground-truth foreground masks (ablation) | G₂ guided segmentation · PASCAL VOC 2012 val, guide strategies without CRF post-processing | mIoU: 51.2→56.9 | Helped |
| Increase GAP seeder output resolution (ablation) | GAP-LowRes · PASCAL VOC 2012 val | mP: 76.5→80.7 | Helped |
| Use dilated convolutions in the GAP seeder (ablation) | GAP-HighRes · PASCAL VOC 2012 val | mP: 80.7→57.7 | Hurt |
| Add CRF loss and post-processing using v1 parameters (ablation) | GAP-HighRes seed-based segmenter · PASCAL VOC 2012 val | mIoU: 37.5→50.4 | Helped |
| Change CRF parameters from v1 to v2 (ablation) | GAP-HighRes seed-based segmenter · PASCAL VOC 2012 val | mIoU: 50.4→45.2 | Hurt |
| Use saliency-guided G₂ instead of the best CRF combination (ablation) | GAP-HighRes seed-based segmenter · PASCAL VOC 2012 val | mIoU: 50.4→55.2 | Helped |
| Combine CRF loss with saliency guidance (ablation) | Saliency-only segmentation · PASCAL VOC 2012 val | mIoU | Hurt |
Closest papers by meaning
- Railroad Is Not a Train: Saliency As Pseudo-Pixel Supervision for Weakly Supervised Semantic Segmentation (CVPR 2021 · similarity 0.91)
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.91)
- Weakly-Supervised Semantic Segmentation by Iteratively Mining Common Object Features (CVPR 2018 · similarity 0.90)
- Multi-Evidence Filtering and Fusion for Multi-Label Classification, Object Detection and Semantic Segmentation Based on Weakly Supervised Learning (CVPR 2018 · similarity 0.89)
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.89)
- Non-Salient Region Object Mining for Weakly Supervised Semantic Segmentation (CVPR 2021 · similarity 0.89)
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.