Research map / Detection and segmentation / Weakly supervised detection and segmentation
Expansion and Shrinkage of Localization for Weakly-Supervised Semantic Segmentation
Jinlong Li, ZEQUN JIE, Xu Wang, xiaolin wei, Lin Ma
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 expansion deformable offset sampler (DT_e) trained with loss maximization (ablation) | ESOL · VOC12 train CAM seed | mIoU: 49.1 -> 53.6 | Improved |
| add feature clipping after expansion sampler (ablation) | ESOL · VOC12 train CAM seed | mIoU: 51.8 -> 53.6 | Improved |
| add shrinkage deformable sampler (DT_s) (ablation) | ESOL · VOC12 train CAM seed | mIoU: 50.9 -> 53.6 | Improved |
| add area loss in shrinkage (ablation) | ESOL · VOC12 train CAM seed | mIoU: 52.8 -> 53.6 | Improved |
| add whole Shrinkage stage (DT_s + area loss) after Expansion (ablation) | ESOL · VOC12 train CAM seed | mIoU: 50.1 -> 53.6 | Improved |
| use deformable layer for loss-maximization expansion rather than plain loss maximization (ablation) | Expansion stage · VOC12 train CAM seed | mIoU: 40.3 -> 48.4 baseline (53.6 full) | Improved |
| apply Expansion then Shrinkage to CAM seeds | IRN baseline · VOC12 train | seed mIoU: 48.4 -> 53.6 | Improved |
| replace IRN with ESOL pseudo labels for segmentation training | DeepLab-v2 · COCO val | mIoU: 41.4 -> 42.6 | Improved |
| Shrinkage after Expansion (ablation) | ESOL · VOC12 train | precision: 41.1 -> 67.1 | Improved |
| vary alpha and gamma/mu hyperparameters (ablation) | ESOL · VOC12 | mIoU | No clear effect |
Closest papers by meaning
- Rethinking Class Activation Mapping for Weakly Supervised Object Localization (ECCV 2020 · similarity 0.89)
- Background Activation Suppression for Weakly Supervised Object Localization and Semantic Segmentation (IJCV 2023 · similarity 0.89)
- Weakly Supervised Semantic Segmentation via Progressive Confidence Region Expansion (CVPR 2025 · similarity 0.89)
- Bridging the Gap Between Classification and Localization for Weakly Supervised Object Localization (CVPR 2022 · similarity 0.88)
- Learning Multi-Modal Class-Specific Tokens for Weakly Supervised Dense Object Localization (CVPR 2023 · similarity 0.88)
- Online Refinement of Low-Level Feature Based Activation Map for Weakly Supervised Object Localization (ICCV 2021 · 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.
