# Analysis Artefacts

Generated post-hoc from the test-set prediction JSONs (`results/*_test_preds.json`)
and the ensemble report.

## Contents
- `confusion_matrices/cm_<model>.png` — per-class confusion matrix (counts overlaid on row-normalised heatmap)
- `roc_curves/roc_<model>.png` — one-vs-rest ROC for every class + macro-AUC in title
- `reliability_diagrams/reliability_<model>.png` — 10-bin reliability with ECE in title
- `per_class/per_class_f1.png` — grouped bar chart, all 9 backbones side-by-side
- `per_class/per_class_f1.json` — raw per-class F1 numbers
- `accuracy_ranking.png` — horizontal bar chart of test accuracy
- `ece_summary.json` — Expected Calibration Error per model