雅鲁藏布江全砂碎屑显微图像数据集

河流砂碎屑组分的鉴定和统计是物源分析的关键步骤，传统显微镜鉴定和人工统计过程费时费力，所获得的数据标准不一，质量参差不齐，不同实验室所获得的数据对比性较差。使用机器辅助技术实现碎屑组分自动鉴定是地质学家的夙愿。要实现这一目标，需要专业地质人员拍摄和标记显微图像文件作为训练基础。基于数据公开、共享的原则，作者将前期耗费大量时间和精力所标记的图像数据集发表出来，供感兴趣的地学、计算机等领域研究人员共享。本数据集包含8734个标记的碎屑颗粒的图像和坐标文件，1876张高清砂粒显微图像，120张编号标记底图和2个砂粒成分鉴定表。本数据集可作为机器学习训练集，也可以作为鉴定其他河流砂碎屑组分的参考。

Identification and statistical analysis of clastic components in fluvial sands are critical steps for provenance analysis. Traditional microscopic identification and manual counting processes are time-consuming and labor-intensive, with acquired datasets featuring inconsistent standards, variable quality, and poor comparability across different laboratories. Automating clastic component identification using machine-assisted techniques has long been a long-cherished wish of geologists. To achieve this goal, professional geologists need to capture and label microscopic image files as the training basis. Following the principle of data openness and sharing, the authors have published the image dataset labeled after expending considerable time and effort in the early stage, to be shared by researchers in relevant fields such as geosciences and computer science who are interested in this research. This dataset contains images and coordinate files of 8734 labeled clastic grains, 1876 high-definition microscopic images of sand grains, 120 numbered labeled base maps, and 2 sand grain composition identification tables. This dataset can serve as a training set for machine learning, as well as a reference for identifying clastic components of other fluvial sands.