Research map / Detection and segmentation / Weakly supervised detection and segmentation
Webly Supervised Semantic Segmentation
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 the three-stage web-image training pipeline with two refinement iterations instead of SEC | WebS semantic segmentation · PASCAL VOC 2012 val | mIoU: 50.7 -> 53.4 | Helped |
| Use the three-stage web-image training pipeline with two refinement iterations instead of SEC | WebS semantic segmentation · PASCAL VOC 2012 test | mIoU: 51.7 -> 55.3 | Helped |
| Add the CRF-based refinement stage to initial training (ablation) | WebS semantic segmentation · PASCAL VOC 2012 val | mIoU: 46.4 -> 51.6 | Helped |
| Add a second CRF-based refinement iteration (ablation) | WebS semantic segmentation · PASCAL VOC 2012 val | mIoU: 51.6 -> 53.4 | Helped |
| Add one CRF-based refinement iteration (ablation) | Class-specific SNNs · Object Discovery dataset | mIoU: 58.0 -> 63.2 | Helped |
| Add a second CRF-based refinement iteration (ablation) | Class-specific SNNs · Object Discovery dataset | mIoU: 63.2 -> 64.8 | Helped |
| Increase realistic-image supervision from 2,000 to 10,582 Rk images while keeping 2,000 Wk images (ablation) | WebS-i2 · PASCAL VOC 2012 val | mIoU: 45.9 -> 48.9 | Helped |
| Increase clean-foreground web images from 2,000 to 4,000 while keeping 10,582 Rk images (ablation) | WebS-i2 · PASCAL VOC 2012 val | mIoU: 48.9 -> 51.5 | Helped |
| Increase realistic-image supervision from 2,000 to 6,000 Rk images while keeping 6,807 Wk images (ablation) | WebS-i2 · PASCAL VOC 2012 val | mIoU: 50.6 -> 51.0 | Helped |
| Increase realistic-image supervision from 6,000 to 10,582 Rk images while keeping 6,807 Wk images (ablation) | WebS-i2 · PASCAL VOC 2012 val | mIoU: 51.0 -> 53.4 | Helped |
| Increase clean-foreground web images from 4,000 to 6,807 while keeping 10,582 Rk images (ablation) | WebS-i2 · PASCAL VOC 2012 val | mIoU: 51.5 -> 53.4 | Helped |
| Train class-specific SNNs from web images rather than use Rubinstein et al.'s object segmentation method | SNN-i2 · Object Discovery dataset | mIoU: 57.3 -> 64.8 | Helped |
Closest papers by meaning
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.92)
- Weakly Supervised Semantic Segmentation Using Web-Crawled Videos (CVPR 2017 · similarity 0.91)
- Bootstrapping the Performance of Webly Supervised Semantic Segmentation (CVPR 2018 · similarity 0.91)
- Weakly-Supervised Semantic Segmentation by Iteratively Mining Common Object Features (CVPR 2018 · similarity 0.91)
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.91)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.88)
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.