Pseudo Label
Pseudo-labeling is a semi-supervised learning technique that leverages unlabeled data by using a model's predictions as pseudo-labels to augment training datasets. Current research focuses on improving the accuracy and reliability of these pseudo-labels, addressing issues like class imbalance and noise through methods such as thresholding, contrastive learning, and teacher-student model architectures. This technique is significant because it allows for training high-performing models with limited labeled data, impacting various applications including object detection, image classification, and medical image segmentation.
Papers
Rethinking Data Heterogeneity in Federated Learning: Introducing a New Notion and Standard Benchmarks
Mahdi Morafah, Saeed Vahidian, Chen Chen, Mubarak Shah, Bill Lin
Semi-Supervised Domain Generalization for Cardiac Magnetic Resonance Image Segmentation with High Quality Pseudo Labels
Wanqin Ma, Huifeng Yao, Yiqun Lin, Jiarong Guo, Xiaomeng Li
Active Self-Training for Weakly Supervised 3D Scene Semantic Segmentation
Gengxin Liu, Oliver van Kaick, Hui Huang, Ruizhen Hu
Learning from Future: A Novel Self-Training Framework for Semantic Segmentation
Ye Du, Yujun Shen, Haochen Wang, Jingjing Fei, Wei Li, Liwei Wu, Rui Zhao, Zehua Fu, Qingjie Liu