LION.evaluation

Evaluating and comparing reconstruction methods.

• Evaluation: the standard evaluation of one problem (test data, operator, metrics); evaluation.run(method) gives a test record (per-sample scores with sample IDs, a summary), saved and reused.

• method(x, operator): a model, a model’s file, a solver, a reconstructor or a function as one call, measurements -> images.

• Comparison: several methods on one evaluation, against a baseline, after checking from the models’ records that the comparison is fair (no test patient trained on; the same problem; the same code). comparison.save(folder) writes it for a paper: tables (markdown, LaTeX, CSV), per-sample scores, box plots, differences to the baseline, example images (best/median/worst or chosen samples, zoom crops), and its spec (Comparison.instantiate(folder)).

• LION.evaluation.figures: those figures as functions, in plain matplotlib (only imported there), and their data (per_sample, per_sample_differences) for any other plot.

• baseline(operator): what methods are compared with, the operator’s default reconstruction (FDK for CT, the inverse for an operator with one).

• scores, summary, table: the same, as functions.

Modules

comparison	Comparing reconstruction methods on one evaluation.
evaluation	Evaluating a reconstruction method on a test set.
figures	Figures of test records: score plots and example images, in plain matplotlib (your style sheet and rcParams apply).
methods	Reconstruction methods as one call: measurements -> images.