Research map / Detection and segmentation / Weakly supervised detection and segmentation
Threshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds
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 |
|---|---|---|---|
| Add per-pixel classification loss to CAM + CRF (ablation) | AMN pseudo-mask generation · PASCAL VOC 2012 train pseudo-masks | mIoU: 54.3 -> 62.1 | Helped |
| Add label conditioning to AMN with per-pixel classification loss (ablation) | AMN pseudo-mask generation · PASCAL VOC 2012 train pseudo-masks | mIoU: 62.1 -> 65.3 | Helped |
| Apply label conditioning at layer4 rather than at lower-level layers (ablation) | AMN · PASCAL VOC 2012 train pseudo-masks | mIoU: 48.2; 51.7; 61.2; 62.1 | Helped |
| Apply label conditioning at multiple layers including layer4 (ablation) | AMN · PASCAL VOC 2012 train pseudo-masks | mIoU: 62.1 -> 61.0; 51.4 | Hurt |
| Replace true image-level labels with a one-vector in label conditioning (ablation) | AMN · PASCAL VOC 2012 train pseudo-masks | mIoU: 58.2% -> 58.6% | No clear effect |
| Encode the image-level label vector plus a noise vector (ablation) | AMN · PASCAL VOC 2012 train pseudo-masks | mIoU: 58.2% -> 60.5% | Helped |
| Encode the ground-truth image-level labels for label conditioning (ablation) | AMN · PASCAL VOC 2012 train pseudo-masks | mIoU: 58.2% -> 62.1% | Helped |
| Use AMN instead of RIB for pseudo-mask generation | WSSS · PASCAL VOC 2012 train pseudo-masks | mIoU: 70.6 -> 72.2 | Helped |
| Use AMN instead of RIB for segmentation | WSSS · MS COCO 2014 val | mIoU: 43.8 -> 44.7 | Helped |
| Use AMN instead of RIB for segmentation | WSSS · PASCAL VOC 2012 val | mIoU: 68.3 -> 69.5 | Helped |
| Perturb the global threshold used with CAM or RIB | CAM and RIB pseudo-mask generation · PASCAL VOC 2012 train pseudo-masks | mIoU robustness to threshold | Hurt |
| Use AMN activation manipulation rather than the conventional CAM activation | Pseudo-mask generation under varying thresholds · PASCAL VOC 2012 train pseudo-masks | mIoU robustness to threshold | Helped |
Closest papers by meaning
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.92)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.91)
- Weakly Supervised Semantic Segmentation via Progressive Confidence Region Expansion (CVPR 2025 · similarity 0.91)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.90)
- DuPL: Dual Student with Trustworthy Progressive Learning for Robust 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.