tomo-center v1.0.0

tomo-center v1.0.0

First public release of tomo-center, the reference standalone
implementation of the AI rotation-axis classifier described in:

Tang, S., et al (2026). AI-based rotation-axis
estimation for parallel-beam tomography.

This is the version cited by the paper and used for its benchmark and
stress tests.

What's in the box

• tomo-center find — score a folder of reconstructed slices (one per
  candidate centre) and report the best centre.
• tomo-center train — fine-tune the shipped classifier on
  facility-specific data.
• Pre-trained checkpoint, downloaded separately from
  https://anl.box.com/s/k85a89kyplzd56hnjudw4ergt6ojfhoa.

Relation to tomocupy

The same model and inference pipeline are also integrated into
tomocupy under
--rotation-axis-method ai. This repository packages the classifier on
its own, without the full reconstruction stack, so it can run against
any folder of pre-reconstructed TIFF slices.

License

BSD-3-Clause (UChicago Argonne LLC). See LICENSE.