VoComni

该数据集包含了2万个伪标记的体积数据，用于全监督预训练的分割任务，这些数据是通过模型集成方法生成的。在生成伪标签的过程中，我们排除了预测置信度低的体积数据。该数据集的任务是对20个器官和肿瘤类别进行分割。

This dataset contains 20,000 pseudo-labeled volumetric data generated via a model ensemble method, and is intended for fully-supervised pre-training in segmentation tasks. During the pseudo-label generation process, volumetric data with low prediction confidence were excluded. The task of this dataset is to perform segmentation covering 20 organ and tumor categories.