Research map / Detection and segmentation / Weakly supervised detection and segmentation
Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach
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 |
|---|---|---|---|
| Extend one-step RRM by training Deeplab-v2 on its generated pseudo masks | RRM · PASCAL VOC 2012 val | mIoU: 62.6 -> 66.3 | Helped |
| Extend one-step RRM by training Deeplab-v2 on its generated pseudo masks | RRM · PASCAL VOC 2012 test | mIoU: 62.9 -> 66.5 | Helped |
| Add the dense energy loss to cross-entropy training (ablation) | RRM segmentation branch with RRM masks · PASCAL VOC 2012 val | mIoU: 0.485 -> 0.626 | Helped |
| Add RRM reliable-region mining instead of using CAM masks directly (ablation) | Segmentation branch with CE loss · PASCAL VOC 2012 val | mIoU: 0.461 -> 0.485 | Helped |
| Add RRM reliable-region mining instead of using CAM masks directly (ablation) | Segmentation branch with joint loss · PASCAL VOC 2012 val | mIoU: 0.557 -> 0.626 | Helped |
| Use the dense energy loss alongside cross-entropy loss (ablation) | CAM pseudo-mask segmentation · PASCAL VOC 2012 val | mIoU: 0.461 -> 0.557 | Helped |
| Use a 0.1 labeled-pixel ratio rather than the 0.3–0.4 range (ablation) | RRM with joint loss · PASCAL VOC 2012 val | mIoU: 0.626 -> 0.428 | Hurt |
| Increase the labeled-pixel ratio from 0.4 to 0.8 (ablation) | RRM with joint loss · PASCAL VOC 2012 val | mIoU: 0.626 -> 0.594 | Hurt |
| Label every pixel rather than using a 0.4 labeled-pixel ratio (ablation) | RRM with joint loss · PASCAL VOC 2012 val | mIoU: 0.626 -> 0.582 | Hurt |
| Increase the labeled-pixel ratio from 0.3 to 0.4 (ablation) | RRM with joint loss · PASCAL VOC 2012 val | mIoU: 0.626 -> 0.626 | No clear effect |
Closest papers by meaning
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.93)
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.93)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.92)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.91)
- DuPL: Dual Student with Trustworthy Progressive Learning for Robust Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.90)
- SFC: Shared Feature Calibration in Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.90)
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.