Research map / Label-efficient and robust learning / Semi-supervised learning
Beyond Pixels: Semi-Supervised Semantic Segmentation with a Multi-scale Patch-based Multi-Label Classifier
Prantik Howlader, Srijan Das, Hieu Le, Dimitris Samaras
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 MPMC plug-in to AugSeg, AEL, U2PL, UniMatch | semi-supervised segmentation baselines · PASCAL VOC Classic 1/16 (ResNet-101) | mIoU: 75.2 -> 77.3 (UniMatch) | Improved |
| add MPMC to baselines on Cityscapes | UniMatch · Cityscapes val 1/16 (ResNet-101) | mIoU: 76.6 -> 78.5 | Improved |
| add MPMC to baselines on medical ACDC | UniMatch · ACDC 3 cases | DSC: 88.9 -> 91.8 | Improved |
| add MPMC to baselines on ACDC with more labels | UniMatch · ACDC 7 cases | DSC: 89.9 -> 90.4 | Improved |
| add MPMC on labeled images only (ablation) | basic teacher-student · Cityscapes 1/16 | mIoU: 68.4 -> 70.8 | Improved |
| add MPMC on unlabeled images (adaptive weights) on top of labeled-only MPMC (ablation) | basic teacher-student · Cityscapes 1/16 | mIoU: 70.8 -> 73.1 | Improved |
| use three scales (original, 7x7, 19x19) instead of a single scale (ablation) | MPMC · Cityscapes 1/16 (AugSeg) | mIoU: 74.1 -> 75.3 | Improved |
| plug MPMC in encoder layer 1 instead of decoder end (ablation) | MPMC · PASCAL VOC | mIoU: 75.8 -> 77.3 | Improved |
| set beta of unsupervised multi-label loss to 0.15 instead of 0 (ablation) | MPMC + UniMatch · Cityscapes 1/16 | mIoU: 77.5 -> 78.5 | Improved |
| increase alpha above 0.25 (0.45) (ablation) | MPMC + UniMatch · Cityscapes 1/16 | mIoU: 78.5 -> 76.9 | Hurt |
| add MPMC to ViT-L based AugSeg | AugSeg (Seg-L) · Cityscapes 1/16 | mIoU: 56.6 -> 58.0 | Improved |
| add MPMC (patch-level) in patches with low multi-label accuracy | basic teacher-student · Cityscapes patch analysis | avg mIoU delta: +4.05 (90+) vs -2.20 (0-10) | Mixed |
Closest papers by meaning
- Pseudo-Label Guided Contrastive Learning for Semi-Supervised Medical Image Segmentation (CVPR 2023 · similarity 0.91)
- Semi-Supervised Semantic Segmentation With High- and Low-Level Consistency (TPAMI 2019 · similarity 0.90)
- RankSeg: Adaptive Pixel Classification with Image Category Ranking for Segmentation (ECCV 2022 · similarity 0.89)
- Semi-Supervised Semantic Segmentation under Label Noise via Diverse Learning Groups (ICCV 2023 · similarity 0.89)
- ScaleMatch: Multi-scale Consistency Enhancement for Semi-supervised Semantic Segmentation (AAAI 2025 · similarity 0.87)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.87)
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.
