RefleX: X-ray diffraction images dataset

Image dataset prepared for the RefleX study, described in <em>"Detecting anomalies in X-ray diffraction images using Convolutional Neural Networks"</em><em>.</em> The dataset contains 6311 X-ray diffraction images in 1024x1024 png format (reflex_img_1024_inter_nearest.zip). The repository also contains a file mapping each image to a set of labels (labels.csv) and files describing the assignment of each image to training, validation, and testing sets (labels_train.csv, labels_val.csv, labels_test.csv). The dataset can be used for multi-label classification. Each diffraction image can exhibit any combination of seven classes: Ice ring, Diffuse Scattering, Background Ring, Non-uniform Detector, Loop Scattering, Strong Background, and Artifact.本数据集专为RefleX研究制备，相关描述见于论文《使用卷积神经网络（Convolutional Neural Networks）检测X射线衍射图像中的异常》。该数据集包含6311张分辨率为1024×1024的PNG格式X射线衍射图像，压缩包文件名为reflex_img_1024_inter_nearest.zip。数据集存储库还附带了将每张图像映射至对应标签集的标签文件（labels.csv），以及用于划分训练集、验证集与测试集的样本分配文件（labels_train.csv、labels_val.csv、labels_test.csv）。该数据集可应用于多标签分类任务，每张衍射图像可同时对应7个类别中的任意组合：冰环（Ice ring）、漫散射（Diffuse Scattering）、背景环（Background Ring）、探测器不均匀（Non-uniform Detector）、环散射（Loop Scattering）、强背景（Strong Background）以及伪影（Artifact）。