A dataset for accurate X-ray single photon recognition algorithm based on multidimensional features and positive sample-unlabeled learning

This dataset provides example observations acquired with a CMOS sensor under two exposure times: 50 ms and 100 ms. For each exposure setting, it includes bright-field images, dark-field images, bright-field images processed using multidimensional feature-based photon event selection, and those processed with a conventional thresholding method. All images are stored in .npz format for efficient loading and processing in Python. This dataset can be used for research on single-photon event extraction algorithms, thresholding methods, and the impact of exposure time on detector performance.