Validation status

Component

Status

Evidence

Remaining work

Ray2D preprocessing

Synthetic validated

docs/CHARUCO_IDENTIFICATION.md, notebook 01

Real noisy image benchmark

Non-central Zernike rayfield

Synthetic validated

Plate oracle, CMO oracle (notebook 04, 06)

Real microscope data

Physical model selection

Synthetic validated

6-oracle matrix (CMO_MODEL_SELECTION.md)

Broader catalogue, real instruments

Physical CMO shared-rig

Synthetic validated

CMO oracle recovery (CMO_PHYSICAL_MODEL.md)

Real CMO microscope

Direct vs rayfield mediation

Exploratory

Notebook 08, direct expected-model fit

Symmetric multi-candidate comparison

Zernike compact fallback

Synthetic validated

Uncatalogued Zernike oracle (notebook 07)

Real unknown optics

Key:

  • Synthetic validated — demonstrated on synthetic oracles with controlled noise. Reproducible from committed notebooks and test suite.

  • Exploratory — infrastructure exists and initial results are available, but the experiment is not yet a complete comparative study. Conclusions should be treated as methodological guidance rather than final quantitative proof.

  • Real-data validation — not yet performed. Requires calibrated images from known microscope instruments.

Test status

  • Fast tests: 76 passed, 0 warnings, ~28 s (pytest -m "not slow")

  • Full suite: 113 passed, 0 warnings, ~9 min (pytest -m "")

Reproducibility

# Fast check (~60 s)
python examples/reproduce_docs_results.py --fast

# Full sweep (~15 min)
python examples/reproduce_docs_results.py --full