PaDIS Reproduction¶
The PaDIS reproduction package provides the complete study workflow around the LION-native implementation: dataset preparation, prior training, fixed- validation tuning, reconstruction and unconditional generation, verification, and final table/figure generation. The pages below divide the maintained PaDIS-Reproduction README into task-focused guides, preserving its complete commands, tables, defaults, runtime estimates, and operational cautions.
Install LION, obtain and process LIDC-IDRI, build caches, and provision a backend.
Reproduce diffusion and PnP training, checkpoint selection, W&B logging, and resumption.
Understand every reconstruction method, implementation track, CT experiment, and generation preset.
Inspect final defaults, inheritance, and the validation-only selection procedure.
Run manual or matrix inference, estimate runtimes, resume work, verify outputs, and generate artefacts.
Understand priors, schedules, implementations, and corrected result identifiers.
Locate each pipeline, platform, training, reconstruction, tuning, verification, and reporting entry point.