SELMA3D
收藏资源简介:
SELMA3D数据集由慕尼黑亥姆霍兹中心环境健康研究中心等机构创建,旨在推动3D光片显微镜图像分割的自监督学习研究。该数据集包含35个大尺寸3D图像,每个图像包含超过1000^3体素,以及315个标注的小块图像,用于模型微调和测试。数据集涵盖了多种生物结构,如血管样和斑点样结构。数据来源于小鼠和人类大脑的光片显微镜图像,经过染色和组织透明化处理。数据集的应用领域包括神经科学、免疫学、肿瘤学和心脏病学,旨在解决3D显微镜图像分割中的领域转移问题,提升模型的泛化能力。
The SELMA3D dataset was developed by institutions including the Helmholtz Zentrum München - German Research Center for Environmental Health, with the goal of advancing self-supervised learning research for 3D light-sheet microscopy image segmentation. It contains 35 large-scale 3D images, each with over 1000³ voxels, as well as 315 annotated image patches for model fine-tuning and testing. The dataset covers a variety of biological structures such as vessel-like and spot-like structures. The data is derived from light-sheet microscopy images of mouse and human brains, which have undergone staining and tissue clearing treatments. Its application domains include neuroscience, immunology, oncology, and cardiology, and it is designed to address the domain shift problem in 3D microscopy image segmentation and improve the generalization ability of models.

- 1SELMA3D challenge: Self-supervised learning for 3D light-sheet microscopy image segmentation慕尼黑亥姆霍兹中心环境健康研究中心, 慕尼黑大学, 慕尼黑工业大学, 帝国理工学院, 康奈尔大学, 苏黎世大学, 上海财经大学, BGI研究院, 浙江大学, 多伦多大学, 名古屋大学 · 2025年



