基于C#的图谱文件管理及标签数据集导出软件开发

C#Based Software Development for Graph-file Management and Labeled Datasets Export

为实现铀矿地质遥感领域专业人员对野外采样数据的管理,并提高数据在机器学习领域的应用效率,基于C#执行效率高、Python利于建立深度学习模型的优势,设计开发基于C#的图谱文件管理及机器学习训练集导出软件。结合C#语言与SQL Server数据库开发针对野外采样数据的图谱管理数据库平台,联合主流的机器学习标签数据集存储方式,运用编程手段实现用户定制的标签数据集生成。使用软件检索功能实现数据的选取,结合“导出数据”功能实现标签文件夹的生成;选用“标签数据集导出”功能分别实现光谱标签文件夹数据的按特征波长导出和图片标签文件夹数据的Tfrecords格式转化的功能。利用此软件可帮助用户实现对野外采样数据的优化管理,快速生成用户输入机器学习模型所需的标签数据集,从而提高数据应用效率。

To realize the management of field sampling data by professionals in the field of geological remote sensing and to improve the efficiency of data application in the field of machine learning,C#-based mapping file management and machine learning training set export software have been designed and developed by using the advantages of high execution efficiency of C#and facilitation of building deep learning models by Python.The database platform of atlas management for field sampling data was implemented by using C#combined with SQL Server database,and user-customized labeled dataset generation was realized by using programming means in combination with the research results of mainstream machine learning labeled dataset storage methods.The software search function was used to select data,and the Export Data function was used to generate label folders;the Label Dataset Export function was used to export spectral label folder data by feature wavelength and conversion of image label folder data to Tfrecords format.The software in this paper can help users to optimize the management of field sampling data,and the tag data set generation function can quickly generate the tag data set to be inputted to machine learning models required by users,thus improving the data application efficiency.