Research map / Detection and segmentation / Weakly supervised detection and segmentation
DHR: Dual Features-Driven Hierarchical Rebalancing in Inter- and Intra-Class Regions for Weakly-Supervised Semantic Segmentation
Sanghyun Jo, Fei Pan, In-Jae Yu, Kyungsu Kim
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 OT-based seed initialization (ablation) | MARS baseline pseudo masks · MS COCO train | mIoU: 52.6 -> 54.3 | Improved |
| add CRF refinement to seed initialization (ablation) | OT seed initialization · MS COCO train | mIoU: 54.3 -> 55.1 | Improved |
| add USS balancing for inter-class regions (ablation) | DHR with OT seeds · MS COCO train | mIoU: 55.1 -> 56.8 | Improved |
| add WSS balancing after USS balancing (ablation) | DHR with OT seeds · MS COCO train | mIoU: 56.8 -> 58.6 | Improved |
| add WSS balancing only (without USS balancing) (ablation) | DHR with OT seeds · MS COCO train | mIoU: 55.1 -> 55.3 | No clear effect |
| add full DHR on top of OT seeds (ablation) | WSS pseudo masks · ADE train | mIoU: 30.2 -> 48.1 | Improved |
| add full DHR on top of OT seeds (ablation) | WSS pseudo masks · Pascal Context train | mIoU: 51.3 -> 64.9 | Improved |
| add DHR to a different WSS method (IRNet) vs MARS | IRNet ResNet-50 DeepLabv2 · PASCAL VOC 2012 val | mIoU: 69.8 -> 72.7 | Improved |
| use CAUSE instead of STEGO as USS features (ablation) | DHR on RSEPM · PASCAL VOC 2012 val | mIoU: 79.2 -> 79.6 | Improved |
| vary threshold tau between 0.5 and 0.9 (ablation) | DHR WSS balancing · MS COCO train | mIoU | No clear effect |
| set tau to 1.0 (bypass WSS balancing) (ablation) | DHR · MS COCO train | mIoU | Hurt |
| apply DHR during training | RSEPM · VOC 2012 training time | training time: 10 hours -> 18 hours | Hurt |
Closest papers by meaning
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.92)
- Threshold Matters in WSSS: Manipulating the Activation for the Robust and Accurate Segmentation Model Against Thresholds (CVPR 2022 · similarity 0.92)
- Weakly Supervised Semantic Segmentation via Progressive Confidence Region Expansion (CVPR 2025 · similarity 0.91)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.90)
- Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.90)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 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.
