TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Revisiting Weak-to-Strong Consistency in Semi-Supervised S...

Revisiting Weak-to-Strong Consistency in Semi-Supervised Semantic Segmentation

Lihe Yang, Lei Qi, Litong Feng, Wayne Zhang, Yinghuan Shi

2022-08-21CVPR 2023 1Semi-supervised Change DetectionSemi-Supervised Semantic SegmentationSemantic SegmentationMedical Image AnalysisSemi-supervised Medical Image Segmentation
PaperPDFCode(official)

Abstract

In this work, we revisit the weak-to-strong consistency framework, popularized by FixMatch from semi-supervised classification, where the prediction of a weakly perturbed image serves as supervision for its strongly perturbed version. Intriguingly, we observe that such a simple pipeline already achieves competitive results against recent advanced works, when transferred to our segmentation scenario. Its success heavily relies on the manual design of strong data augmentations, however, which may be limited and inadequate to explore a broader perturbation space. Motivated by this, we propose an auxiliary feature perturbation stream as a supplement, leading to an expanded perturbation space. On the other, to sufficiently probe original image-level augmentations, we present a dual-stream perturbation technique, enabling two strong views to be simultaneously guided by a common weak view. Consequently, our overall Unified Dual-Stream Perturbations approach (UniMatch) surpasses all existing methods significantly across all evaluation protocols on the Pascal, Cityscapes, and COCO benchmarks. Its superiority is also demonstrated in remote sensing interpretation and medical image analysis. We hope our reproduced FixMatch and our results can inspire more future works. Code and logs are available at https://github.com/LiheYoung/UniMatch.

Results

TaskDatasetMetricValueModel
Medical Image SegmentationACDC 5% labeled dataDice (Average)87.61UniMatch
Medical Image SegmentationACDC 10% labeled dataDice (Average)89.92UniMatch
Medical Image SegmentationACDC 20% labeled dataDice (Average)90.47UniMatch
Semantic SegmentationCOCO 1/512 labeledValidation mIoU31.9UniMatch
Semantic SegmentationCOCO 1/256 labeledValidation mIoU38.9UniMatch
Semantic SegmentationADE20K 1/16 labeledValidation mIoU31.5UniMatch
Semantic SegmentationPascal VOC 2012 6.25% labeledValidation mIoU80.94UniMatch (DeepLab v3+ with ResNet-101)
Semantic SegmentationPASCAL VOC 2012 92 labeledValidation mIoU75.2UniMatch (DeepLab v3+ with ResNet-101)
Semantic SegmentationADE20K 1/32 labeledValidation mIoU28.1UniMatch
Semantic SegmentationPASCAL VOC 2012 732 labeledValidation mIoU79.9UniMatch (DeepLab v3+ with ResNet-101)
Semantic SegmentationPASCAL VOC 2012 1464 labelsValidation mIoU81.2UniMatch (DeepLab v3 with ResNet-101)
Semantic SegmentationPASCAL VOC 2012 25% labeledValidation mIoU80.43UniMatch (DeepLab v3+ with ResNet-101)
Semantic SegmentationCOCO 1/128 labeledValidation mIoU44.5UniMatch
Semantic SegmentationCOCO 1/64 labeledValidation mIoU48.2UniMatch
Semantic SegmentationCityscapes 100 samples labeledValidation mIoU73UniMatch (DeepLab v3+ with ResNet-101)
Semantic SegmentationPASCAL VOC 2012 366 labeledValidation mIoU78.8UniMatch (DeepLab v3+ with ResNet-101)
Semantic SegmentationCityscapes 6.25% labeledValidation mIoU76.59UniMatch (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)
Semantic SegmentationCOCO 1/32 labeledValidation mIoU49.8UniMatch
Semantic SegmentationPASCAL VOC 2012 183 labeledValidation mIoU77.2UniMatch (DeepLab v3+ with ResNet-101)
Change DetectionWHU - 20% labeled dataIoU81.7UniMatch
Change DetectionWHU - 5% labeled dataIoU80.2UniMatch
Change DetectionWHU - 10% labeled dataIoU81.7UniMatch
Change DetectionWHU - 40% labeled dataIoU85.1UniMatch
Change DetectionLEVIR-CD - 10% labeled dataIoU82UniMatch
Change DetectionLEVIR-CD - 5% labeled dataIoU80.7UniMatch
Change DetectionLEVIR-CD - 20% labeled dataIoU81.7UniMatch
Change DetectionLEVIR-CD - 40% labeled dataIoU82.1UniMatch
10-shot image generationCOCO 1/512 labeledValidation mIoU31.9UniMatch
10-shot image generationCOCO 1/256 labeledValidation mIoU38.9UniMatch
10-shot image generationADE20K 1/16 labeledValidation mIoU31.5UniMatch
10-shot image generationPascal VOC 2012 6.25% labeledValidation mIoU80.94UniMatch (DeepLab v3+ with ResNet-101)
10-shot image generationPASCAL VOC 2012 92 labeledValidation mIoU75.2UniMatch (DeepLab v3+ with ResNet-101)
10-shot image generationADE20K 1/32 labeledValidation mIoU28.1UniMatch
10-shot image generationPASCAL VOC 2012 732 labeledValidation mIoU79.9UniMatch (DeepLab v3+ with ResNet-101)
10-shot image generationPASCAL VOC 2012 1464 labelsValidation mIoU81.2UniMatch (DeepLab v3 with ResNet-101)
10-shot image generationPASCAL VOC 2012 25% labeledValidation mIoU80.43UniMatch (DeepLab v3+ with ResNet-101)
10-shot image generationCOCO 1/128 labeledValidation mIoU44.5UniMatch
10-shot image generationCOCO 1/64 labeledValidation mIoU48.2UniMatch
10-shot image generationCityscapes 100 samples labeledValidation mIoU73UniMatch (DeepLab v3+ with ResNet-101)
10-shot image generationPASCAL VOC 2012 366 labeledValidation mIoU78.8UniMatch (DeepLab v3+ with ResNet-101)
10-shot image generationCityscapes 6.25% labeledValidation mIoU76.59UniMatch (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)
10-shot image generationCOCO 1/32 labeledValidation mIoU49.8UniMatch
10-shot image generationPASCAL VOC 2012 183 labeledValidation mIoU77.2UniMatch (DeepLab v3+ with ResNet-101)

Related Papers

SeC: Advancing Complex Video Object Segmentation via Progressive Concept Construction2025-07-21DiffOSeg: Omni Medical Image Segmentation via Multi-Expert Collaboration Diffusion Model2025-07-17SCORE: Scene Context Matters in Open-Vocabulary Remote Sensing Instance Segmentation2025-07-17Unified Medical Image Segmentation with State Space Modeling Snake2025-07-17A Privacy-Preserving Semantic-Segmentation Method Using Domain-Adaptation Technique2025-07-17MUPAX: Multidimensional Problem Agnostic eXplainable AI2025-07-17SAMST: A Transformer framework based on SAM pseudo label filtering for remote sensing semi-supervised semantic segmentation2025-07-16Tomato Multi-Angle Multi-Pose Dataset for Fine-Grained Phenotyping2025-07-15