An overview of the pipeline proposed here. Raw X-ray image volumes (left) are iteratively segmented through collaborative human-machine teaming. Machine-guided annotation in our web application yields training data for the segmentation models. The segmentation model is applied to the whole dataset through batch processing, taking advantage of chunked data storage paradigms. This process can be repeated until the model meets human-defined quality benchmarks. Finally, a high-resolution segmentation mask can be exported, alongside image renders and meshes suitable for rendering in 3D software.

 
 
  Part of: During MAD, Matelsky JK, Gustafsson FK, Voeten DFAE, Chen D, Wester BA, Kording KP, Ahlberg PE, Schön TB (2025) Automated segmentation of synchrotron-scanned fossils. Fossil Record 28(1): 103-114. https://doi.org/10.3897/fr.28.e139379