CVPR2022
- A Dual Weighting Label Assignment Scheme for Object Detection
一种用于目标检测的双加权标签分配方案
香港理工大学
双加权的自适应标签分配方案，打破耦合加权的惯例，从不同方面估计一致性和不一致性度量，为每个锚点动态分配单独的正负权重，训练精确的密集目标检测器
新的方框细化操作——直接细化回归图上的方框
但是，会减少训练样本的数量，影响对 small objects 的训练效果
- ACPL: Anti-curriculum Pseudo-labelling for Semi-supervised Medical Image Classification
ACPL：用于半监督医学图像分类的反课程伪标记
澳大利亚机器学习研究所，阿德莱德大学，德国乌尔姆第二大学
反课程伪标签的 SSL 方法
新的伪标记方法 CSDI —— 不需要估计类分类阈值
IM 集合 KNN 分类器，生成更准确的伪标签
ASP 方法保持密度评分在后续训练阶段的准确性
https://github.com/FBLADL/ACPL
- ADeLA_ Automatic Dense Labeling With Attention for Viewpoint Shift in Semantic Segmentation
语义分割中注意力视点转移的自动密集标记
斯坦福、谷歌
将视图转换网络视为（彩色图像隐含的未知映射的）函数表示（a functional representation of an unknown mapping implied by the color images）
提出函数标签幻觉法（functional label hallucination）（在目标域中生成具有不确定性的伪标签）
- Back to Reality_ Weakly-Supervised 3D Object Detection With Shape-Guided Label Enhancement
回归现实：基于形状导向标签增强的弱监督 3D 物体检测
清华大学、谷歌
新的标签增强方法 —— BR（Back to Reality），用于只使用对象中心和类标签作为监督训练的 3D 对象检测
https://github.com/wyf-ACCEPT/BackToReality
- BoostMIS_ Boosting Medical Image Semi-Supervised Learning With Adaptive Pseudo Labeling and Informative Active Annotation
MoostMIS：使用自适应伪标记和信息性主动性标注促进医学图像半监督学习
新加坡国立大学医学研究院
新的医学图像半监督框架 —— BoostMIS，利用主动学习释放未标记数据的潜力
提出自适应伪标记和信息主动标记，形成闭环
自适应伪标记 —— 动态调整阈值
https://github.com/wannature/BoostMIS
- Cross-Model Pseudo-Labeling for Semi-Supervised Action Recognition
跨模型伪标记的半监督动作识别
香港中文大学，南洋理工大学，微软亚洲研究院
新的半监督动作识别框架 —— 跨模型伪标记框架 CMPL
将轻量级网络与主干网络配对，通过伪标记相互学习
- DASO_ Distribution-Aware Semantics-Oriented Pseudo-Label for Imbalanced Semi-Supervised Learning
面向分布感知语义导向的非平衡半监督学习伪标签
韩国科学技术院
通用的伪标记框架，将 “基于相似性的分类器中的语义伪标签” 与 “线性分类器中的线性伪标签” 进行类自适应混合
https://github.com/ytaek-oh/daso
- Debiased Learning From Naturally Imbalanced Pseudo-Labels
自然不平衡伪标签的去偏学习
加州大学伯克利分校，微软研究院
通用附加组件 —— 去偏伪标记方案，在不利用任何真实数据分布的先验知识的情况下，动态地减轻偏误伪标签对学生模型的影响
可以集成到半监督学习 SSL 和零镜头学习 ZSL 中的模块
不平衡、长尾分布
https://github.com/frank-xwang/debiased-pseudo-labeling
- Deep Anomaly Discovery From Unlabeled Videos via Normality Advantage and Self-Paced Refinement
- Dist-PU_ Positive-Unlabeled Learning From a Label Distribution Perspective
- Exploiting Pseudo Labels in a Self-Supervised Learning Framework for Improved Monocular Depth Estimation
- FedCorr_ Multi-Stage Federated Learning for Label Noise Correction
- Few-Shot Incremental Learning for Label-to-Image Translation
- Few-Shot Learning With Noisy Labels
带噪声标签的少样本学习
Facebook AI 研究院，麦吉尔大学
transformer
- Improving Segmentation of the Inferior Alveolar Nerve Through Deep Label Propagation
基于深度标签传播的牙槽神经分割改进
标签传播技术：稀疏注释扩展为密集标签
- Incorporating Semi-Supervised and Positive-Unlabeled Learning for Boosting Full Reference Image Quality Assessment
结合半监督和正无标记学习促进全参考图像质量评估
https://github.com/happycaoyue/JSPL
- Incremental Learning in Semantic Segmentation From Image Labels
增量学习在图像标签语义分割中的应用
https://github.com/fcdl94/WILSON
- Instance-Dependent Label-Noise Learning With Manifold-Regularized Transition Matrix Estimation
基于流形正则化过渡矩阵估计的实例相关标签噪声学习
标签噪声学习
过渡矩阵
统计一致性分类器
- Interactive Multi-Class Tiny-Object Detection
交互式多类小目标检测
https://github.com/ChungYi347/Interactive-Multi-Class-Tiny-Object-Detection
- Label Matching Semi-Supervised Object Detection
标签匹配半监督对象检测
伪标签，均值教师、自适应标签分布
https://github.com/HIK-LAB/SSOD
- Label Relation Graphs Enhanced Hierarchical Residual Network for Hierarchical Multi-Granularity Classification
面向分层多粒度分类的标签关系图增强分层残差网络
https://github.com/MonsterZhZh/HRN
- Label, Verify, Correct_ A Simple Few Shot Object Detection Method
标记、验证、校正:一种简单的少样本物体检测方法
高质量伪标签
COCO
VOC
物体检测、半监督、少样本
- Label-Only Model Inversion Attacks via Boundary Repulsion
基于边界排斥的标签模型反转攻击
https://github.com/m-kahla/Label-Only-Model-Inversion-Attacks-via-Boundary-Repulsion
- Large Loss Matters in Weakly Supervised Multi-Label Classification
弱监督多标签分类中的大损失问题
partial label setting
learn a multi-label classification using partially observed labels per image
the model first learns the representation of clean labels, and then starts memorizing noisy labels
https://github.com/snucml/LargeLossMatters
- Large-Scale Pre-training for Person Re-identification with Noisy Labels
带噪声标签的人再识别的大规模预训练
https://github.com/DengpanFu/LUPerson-NL
- Learning Fair Classifiers With Partially Annotated Group Labels
学习带有部分注释的组标签的公平分类器
https://github.com/naver-ai/cgl_fairness
- Learning From Pixel-Level Noisy Label_ A New Perspective for Light Field Saliency Detection
从像素级噪声标签学习:光场显著性检测的新视角
https://github.com/OLobbCode/NoiseLF
- Learning To Detect Mobile Objects From LiDAR Scans Without Labels
学习从无标签的激光雷达扫描中检测移动物体
https://github.com/YurongYou/MODEST
- Learning To Imagine_ Diversify Memory for Incremental Learning Using Unlabeled Data
学习想象:使用无标签数据进行增量学习的多样化记忆
语义对比学习
- Learning With Neighbor Consistency for Noisy Labels
噪声标签的邻居一致性学习
领域一致性正则化：具有标签噪声的深度学习策略
半监督、转导标签传播
- Learning With Twin Noisy Labels for Visible-Infrared Person Re-Identification
双噪声标签学习的可见-红外人员再识别
https://github.com/XLearning-SCU/2022-CVPR-DART
- Multi-class Token Transformer for Weakly Supervised Semantic Segmentation
- Multidimensional Belief Quantification for Label-Efficient Meta-Learning
- Multi-Label Classification With Partial Annotations Using Class-Aware Selective Loss
- Multi-Label Iterated Learning for Image Classification With Label Ambiguity
- Multi-Marginal Contrastive Learning for Multi-Label Subcellular Protein Localization
- Mutual Quantization for Cross-Modal Search With Noisy Labels
- Not All Labels Are Equal_ Rationalizing the Labeling Costs for Training Object Detection
- Not All Relations Are Equal_ Mining Informative Labels for Scene Graph Generation
- On Learning Contrastive Representations for Learning With Noisy Labels
- Open-Vocabulary Instance Segmentation via Robust Cross-Modal Pseudo-Labeling
- Part-Based Pseudo Label Refinement for Unsupervised Person Re-Identification
- PLAD_ Learning To Infer Shape Programs With Pseudo-Labels and Approximate Distributions
- PNP_ Robust Learning From Noisy Labels by Probabilistic Noise Prediction
- Propagation Regularizer for Semi-Supervised Learning With Extremely Scarce Labeled Samples
- Replacing Labeled Real-Image Datasets With Auto-Generated Contours
- Safe-Student for Safe Deep Semi-Supervised Learning With Unseen-Class Unlabeled Data
- Scalable Penalized Regression for Noise Detection in Learning With Noisy Labels
- Selective-Supervised Contrastive Learning With Noisy Labels
- Self-Supervised Global-Local Structure Modeling for Point Cloud Domain Adaptation With Reliable Voted Pseudo Labels
- Self-Taught Metric Learning without Labels
- Semi-Supervised Learning of Semantic Correspondence With Pseudo-Labels
- Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels
- The Devil Is in the Labels_ Noisy Label Correction for Robust Scene Graph Generation
- The Devil Is in the Margin_ Margin-Based Label Smoothing for Network Calibration
- The Neurally-Guided Shape Parser_ Grammar-Based Labeling of 3D Shape Regions With Approximate Inference
- Towards Data-Free Model Stealing in a Hard Label Setting
- TWIST_ Two-Way Inter-Label Self-Training for Semi-Supervised 3D Instance Segmentation
- Undoing the Damage of Label Shift for Cross-Domain Semantic Segmentation
- UniCon_ Combating Label Noise Through Uniform Selection and Contrastive Learning
- Unified Contrastive Learning in Image-Text-Label Space
- Unimodal-Concentrated Loss_ Fully Adaptive Label Distribution Learning for Ordinal Regression
- Use All The Labels_ A Hierarchical Multi-Label Contrastive Learning Framework
- Which Images To Label for Few-Shot Medical Landmark Detection_
