# SSL4MIS

**Repository Path**: he-junyang/SSL4MIS

## Basic Information

- **Project Name**: SSL4MIS
- **Description**: SSL4MIS仓库，写注释用于自我学习
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-07-17
- **Last Updated**: 2025-07-30

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

SSL4MIS项目地址：https://gitcode.com/gh_mirrors/ss/SSL4MIS

（1）Mean Teacher：Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results 2017

（2）Entropy Minimization：ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation 2018

（3）Deep Adversarial Networks：Deep Adversarial Networks for Biomedical Image Segmentation Utilizing Unannotated Images 2017

（4）Uncertainty Aware Mean Teacher：Uncertainty-aware Self-ensembling Model for Semi-supervised 3D Left Atrium Segmentation 2019

（5）Interpolation Consistency Training：Interpolation consistency training for semi-supervised learning 2022

（6）Uncertainty Rectified Pyramid Consistency：Efficient Semi-Supervised Gross Target Volume of Nasopharyngeal Carcinoma Segmentation via Uncertainty Rectified Pyramid Consistency 2020

（7）Cross Pseudo Supervision：Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision 2021

（8）Cross Consistency Training：Semi-Supervised Semantic Segmentation with Cross-Consistency Training 2020

（9）Deep Co-Training：Deep Co-Training for Semi-Supervised Image Recognition 2018

（10）Cross Teaching between CNN and Transformer：Semi-Supervised Medical Image Segmentation via Cross Teaching between CNN and Transformer 2021

（11）FixMatch：FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence 2020

（12）Regularized Dropout：R-Drop: Regularized Dropout for Neural Networks 2021 

半监督算法的成立离不开四类假设：平滑性假设、聚类假设、流形假设、低密度分离假设

因为平滑性说明两个样本在特征空间映射的点是接近的，那么它们对应的属性值也是接近的；聚类假设假设数据存在簇（聚类）结构，那么在同一个簇内的样本应该属于同一类。平滑性假设和聚类假设其实是定量和定性的关系。直观来说，一个物体和背景，它们的类别索引在边界上离得比较远，在物体和背景内部的类别索引变化为0，对应着分类边界平滑性大、目标内部的平滑性小。