view article
Figure 1
Overview of SMAXI's architecture. Multidimensional raw X-ray images, ranging from static 2D X-ray images to 3D time-resolved in situ CT data can be utilized as input data. These data are then processed through a sequential workflow: (1) image pre-processor to enhance the image contrast through pixel normalization, (2) SAM-based object segmentation feature for annotating 2D and 3D objects, (3) YOLO-driven object detection and trajectory tracking, and (4) LLM-assisted geometrical feature analysis. Finally, the morphological features of tracked objects, such as area, perimeter, width, depth, and aspect ratio, are quantified as output.

Journal logoJOURNAL OF
SYNCHROTRON
RADIATION
ISSN: 1600-5775
Volume 33| Part 5| September 2026|
https://doi.org/10.1107/S1600577526007745