Luddy School of Informatics, Computing, and Engineering, Indiana University（印第安纳大学卢迪信息、计算与工程学院）；Tulane University（杜兰大学）；DEVCOM Army Research Laboratory（美国陆军作战能力发展司令部陆军研究实验室）
Zheng Chen, Zhengming Ding, Jason M. Gregory, Lantao Liu
Mixup-based data augmentation has been validated to be a critical stage in the self-training framework for unsupervised domain adaptive semantic segmentation (UDA-SS), which aims to transfer knowledge from a well-annotated (source) domain to an unlabeled (target) domain. Existing self-training methods usually adopt the popular region-based mixup techniques with a random sampling strategy, which unfortunately ignores the dynamic evolution of different semantics across various domains as training proceeds. To improve the UDA-SS performance, we propose an Informed Domain Adaptation (IDA) model, a self-training framework that mixes the data based on class-level segmentation performance, which aims to emphasize small-region semantics during mixup. In our IDA model, the class-level performance is tracked by an expected confidence score (ECS). We then use a dynamic schedule to determine the mixing ratio for data in different domains. Extensive experimental results reveal that our proposed method is able to outperform the state-of-the-art UDA-SS method by a margin of 1.1 mIoU in the adaptation of GTA-V to Cityscapes and of 0.9 mIoU in the adaptation of SYNTHIA to Cityscapes.
Language-guided Medical Image Segmentation with Target-informed Multi-level Contrastive Alignments
School of Computer Science, The University of Sydney（悉尼大学计算机科学学院）；Institute of Translational Medicine, Shanghai Jiao Tong University（上海交通大学转化医学研究院）；Zhongguancun Academy & Zhongguancun Institute of Artificial Intelligence, Beijing, China（北京中关村学院及中关村人工智能研究院）；Department of Molecular Imaging, Royal Prince Alfred Hospital（皇家珀斯阿尔弗雷德医院分子成像部）
Understanding-informed Bias Mitigation for Fair CMR Segmentation
College of Electronic and Information Engineering, Tongji University（同济大学电子与信息工程学院）；Guy’s and St Thomas’ NHS Foundation Trust（盖茨和圣托马斯国家健康服务基金会信托）；HeartFlow Inc, United States（美国HeartFlow公司）；British Heart Foundation Centre, Department of Clinical Pharmacology, King’s College London（英国心脏基金会中心、伦敦国王学院临床药理学系）；School of Biomedical Engineering & Imaging Sciences, King’s College London（伦敦国王学院生物医学工程与成像科学学院）
Plume Segmentation from MethaneSAT with Cross-Sensor Transfer Learning and Physics-Informed Postprocessing
Center for Astrophysics | Harvard & Smithsonian（哈佛-史密松天体物理中心）；Environmental Defense Fund（环境防御基金）；Department of Earth and Planetary Sciences, Harvard University（哈佛大学地球与行星科学系）；Institute of Environmental Physics (IUP), University of Bremen（不莱梅大学环境物理研究所）；John A. Paulson School of Engineering and Applied Sciences, Harvard University（哈佛大学约翰·A·保罗森工程与应用科学学院）
Less is More: Reducing Task and Model Complexity for 3D Point Cloud Semantic Segmentation
Durham University（杜伦大学）
虽然近年来3D LiDAR点云数据的可用性显著增长，但注释仍然昂贵且耗时，导致对自动驾驶等应用领域的半监督语义分割方法的需求。现有的工作往往采用相对较大的分割骨干网络，以提高分割精度，在计算成本为代价。此外，许多人使用均匀采样来减少学习所需的地面实况数据要求，通常导致次优性能。为了解决这些问题，我们提出了一个新的管道，采用更小的架构，需要更少的地面实况注释，以实现卓越的分割精度相比，当代的方法。这是通过一个新的稀疏依赖可分离卷积模块来实现的，该模块在保持整体任务性能的同时显著减少了网络参数计数。为了有效地对我们的训练数据进行子采样，我们提出了一种新的时空冗余帧下采样（ST-RFD）方法，该方法利用环境中传感器运动的知识来提取训练数据帧样本的更多样化的子集。为了利用有限的注释数据样本的使用，我们进一步提出了一种由LiDAR反射率通知的软伪标签方法。在SemanticKITTI（59.5@5%）和ScribbleKITTI（58.1@5%）基准数据集上，我们的方法在mIoU方面优于当代半监督工作，基于2.3倍的模型参数减少和641倍的乘加运算，同时还在有限的训练数据（即，Less is More）。
Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label
Institute of Medical University, Peking University（北京大学医学部）；National Biomedical Imaging Center, Peking University（北京大学生物医学成像中心）；Institute of Biomedical Engineering, Peking University Shenzhen Graduate School（北京大学深圳研究生院生物医学工程研究所）
Post-Hurricane Debris Segmentation Using Fine-Tuned Foundational Vision Models
Department of Civil and Environmental Engineering, Rice University（土木与环境工程系， Rice大学）；Department of Electrical and Computer Engineering, Rice University（电气与计算机工程系， Rice大学）；Ken Kennedy Institute, Rice University（肯尼迪研究所， Rice大学）
Uncertainty Quantification in Medical Image Segmentation with Multi-decoder U-Net
Harbin Institute of Technology at Shenzhen（哈尔滨工业大学（深圳））；Peng Cheng Lab（鹏城实验室）；MindsGo Co.Ltd（MindsGo 有限公司）；Capital Medical University（首都医科大学）；Xuanwu Hospital Capital Medical University（首都医科大学宣武医院）
Generalizing Abstention for Noise-Robust Learning in Medical Image Segmentation
Institute of Computer Science, University of Bonn（博恩大学计算机科学研究所）；Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS（弗劳恩霍夫智能分析与信息系统研究所）；Lamarr Institute for Machine Learning and Artificial Intelligence（拉马尔人工智能与机器学习研究所）
Sequential 3D U-Nets for Biologically-Informed Brain Tumor Segmentation
Department of Neuro-Oncology, Massachusetts General Hospital（麻省总医院神经肿瘤科）；MGH & BWH Center for Clinical Data Science（麻省总医院与布莱根妇女医院临床数据科学中心）；Department of Radiology, Massachusetts General Hospital（麻省总医院放射科）；Massachusetts General Hospital（麻省总医院）；NVIDIA（英伟达）
Uncertainty-informed Mutual Learning for Joint Medical Image Classification and Segmentation
Sichuan University（四川大学）；West China Hospital, Sichuan University（四川大学华西医院）；Institute of High Performance Computing, Agency for Science, Technology and Research（新加坡科技研究局高性能计算研究所）
