Research map / Label-efficient and robust learning / Semi-supervised learning
Logic-induced Diagnostic Reasoning for Semi-supervised Semantic Segmentation
Chen Liang, Wenguan Wang, Jiaxu Miao, Yi Yang
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 hierarchical prediction head (ablation) | DeepLabV3+ consistency regularization baseline · VOC aug 1/16 | mIoU: 68.02 -> 69.15 | Improved |
| add minimal diagnosis computation with uniform sampling of conflicts (ablation) | hierarchical prediction · VOC aug 1/16 | mIoU: 69.15 -> 71.83 | Improved |
| sample diagnoses from fuzzy diagnosis likelihood (ablation) | LogicDiag · VOC aug 1/16 | mIoU: 71.83 -> 73.25 | Improved |
| add composition, decomposition and exclusion rules together (ablation) | LogicDiag · VOC aug 1/16 | mIoU: 69.15 -> 73.25 | Improved |
| add exclusion rule only (ablation) | LogicDiag · VOC aug 1/16 | mIoU: 69.15 -> 71.31 | Improved |
| use Monte Carlo sampling instead of uniform/predictive/greedy diagnosis resolution (ablation) | LogicDiag · VOC aug 1/16 | mIoU: 72.61 -> 73.25 (vs Greedy) | Improved |
| use random label hierarchy instead of official hierarchy (ablation) | LogicDiag · VOC aug 1/16 | mIoU: 73.25 -> 67.11 | Hurt |
| add LogicDiag on top of MKD | MKD · VOC original 1/16 | mIoU: 73.25 -> 74.70 | Improved |
| add LogicDiag logic diagnosis vs prior SOTA | semi-supervised segmentation · COCO val | mIoU: 47.25 -> 50.51 (1/32 vs MKD) | Improved |
| run logic diagnosis during training | LogicDiag · VOC aug 1/16 | training time (min per 10K iters): 85.3 -> 89.3 | No clear effect |
Closest papers by meaning
- LogicSeg: Parsing Visual Semantics with Neural Logic Learning and Reasoning (ICCV 2023 · similarity 0.90)
- Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels (CVPR 2022 · similarity 0.89)
- PseudoSeg: Designing Pseudo Labels for Semantic Segmentation (ICLR 2021 · similarity 0.89)
- GuidedMix-Net: Semi-supervised Semantic Segmentation by Using Labeled Images as Reference (AAAI 2022 · similarity 0.89)
- DAW: Exploring the Better Weighting Function for Semi-supervised Semantic Segmentation (NeurIPS 2023 · similarity 0.88)
- Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant (NeurIPS 2022 · 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.
