BMz galaxies
收藏资源简介:
BMz galaxies数据集是由DESI Legacy Imaging Surveys中的BASS和MzLS调查所产生的,包含248,088个星系的详细形态分类。该数据集通过无监督域适应方法从DECaLS数据集中迁移学习而来,旨在解决不同调查间信号噪声比和分辨率的差异问题。数据集的创建过程包括对DECaLS图像的预训练模型进行微调,并使用BMz的未标记数据进行进一步训练。该数据集主要应用于星系形态分类研究,旨在提高对不同调查中星系形态预测的准确性。
The BMz galaxies dataset is generated from the BASS and MzLS surveys within the DESI Legacy Imaging Surveys, containing detailed morphological classifications for 248,088 galaxies. It adopts unsupervised domain adaptation-based transfer learning from the DECaLS dataset, aiming to address the discrepancies in signal-to-noise ratio and resolution across different surveys. The development of this dataset involves fine-tuning a pre-trained model on DECaLS images, followed by further training using the unlabeled data from BMz. This dataset is primarily applied to research on galaxy morphological classification, with the goal of improving the accuracy of galaxy morphology predictions across different surveys.




