Deep learning–driven chromatic X‐ray imaging based on multicolor halide scintillation film stacks for quantitative densitometry

Abstract X‐ray imaging is foundational and urgent to biomedical diagnosis and industrial nondestructive inspections; however, conventional single‐dose approaches offer limited material specificity. Here, we present a deep learning–driven X‐ray imaging technique leveraging vapor‐deposited multicolor halide scintillation film stacks, enabling imaging and the systematic curation of a materials genome database. By coupling an Attention U‐Net with extreme Gradient Boosting, we further develop an end‐to‐end density regression system that outputs quantitative material information directly from chromatic images, realizing segmentation of circuit‐ and pixel‐level mapping from red/green/blue (RGB) values to densities, with an intersection over union of 81% and a root mean square error of ±0.622 g cm −3 . The effectiveness has been demonstrated through segmentation of complex circuit boards and 3D reconstruction of material density distributions. This work delivers a reliable and intelligent material platform and technique for chromatic X‐ray imaging, offering enhanced material discrimination and nondestructive testing capabilities.