Research map / Detection and segmentation / Weakly supervised detection and segmentation
Escaping the CAM Shadow: Uncertainty-Guided Reliable Learning for Weakly Supervised Semantic Segmentation
Luyao Chang, Leiting Chen, Chen Yang, Chuan Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 class-wise uncertainty modulation of classification loss (PUM-c) (ablation) | MiT-B1 baseline · PASCAL VOC 2012 val | segmentation mIoU (%): 63.4 -> 68.2 | Improved |
| propagate uncertainty to pixel-level segmentation loss (PUM-p) (ablation) | UGRL setting I · PASCAL VOC 2012 val | segmentation mIoU (%): 68.2 -> 70.4 | Improved |
| add contrastive learning on decoder features (ablation) | UGRL setting II · PASCAL VOC 2012 val | segmentation mIoU (%): 70.4 -> 74.0 | Improved |
| add uncertainty guidance to contrastive learning (full UGRL) (ablation) | UGRL setting III · PASCAL VOC 2012 val | segmentation mIoU (%): 74.0 -> 77.4 | Improved |
| add hyperdimensional projection for prototypes (ablation) | UGRL prototype space · VOC prototype similarity | class-prototype separation | Improved |
| increase reliable pool K from 35% to 80% (ablation) | UGRL · PASCAL VOC 2012 val | mIoU (%): 77.4 -> 74.2 | Hurt |
| decrease reliable pool K from 35% to 5% (ablation) | UGRL · MS COCO 2014 val | mIoU (%): 48.2 -> 44.1 | Hurt |
| increase H_cls dimension beyond 1024 (ablation) | UGRL · PASCAL VOC 2012 | mIoU (%) | Hurt |
| use UGRL vs prior best single-stage methods | MiT-B1 · PASCAL VOC 2012 val | mIoU (%): 75.5 -> 77.4 | Improved |
Closest papers by meaning
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.91)
- DuPL: Dual Student with Trustworthy Progressive Learning for Robust Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.90)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.89)
- UNICL-SAM: Uncertainty-Driven In-Context Segmentation with Part Prototype Discovery (CVPR 2025 · similarity 0.89)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · 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.
