Document analysis/, gradcam/, and kfold/ artefacts in README

README.md

@@ -338,3 +338,28 @@ Apache-2.0 for code and weights. Original Mendeley dataset retains its
 own licence (CC BY 4.0).
 
 Questions / collaboration: open an issue on the Hub repo.
+
+
+## Visual Analysis & Supporting Material
+
+In addition to the headline numbers above, the repo ships:
+
+### `analysis/`
+- **`confusion_matrices/cm_<model>.png`** — per-class confusion matrix for each of the 9 models + soft-vote ensemble (10 PNGs)
+- **`roc_curves/roc_<model>.png`** — one-vs-rest ROC for every class with macro-average AUC (10 PNGs)
+- **`reliability_diagrams/reliability_<model>.png`** — 10-bin reliability diagram + Expected Calibration Error (ECE)
+- **`per_class/per_class_f1.png`** + `.json` — grouped bar chart of per-class F1 across all 9 backbones
+- **`accuracy_ranking.png`** — horizontal bar chart of test accuracy (best on top)
+- **`ece_summary.json`** — ECE values for every model
+
+### `gradcam/`
+Per-class Grad-CAM saliency maps (one PNG per CNN-class model, 10 classes per figure) for VGG19, ResNet50, ResNet101, DenseNet121, InceptionV3, and CLIP. Generated with `pytorch-grad-cam` against the last convolutional block of each backbone (input-gradient saliency for CLIP).
+
+### `kfold/`
+5-fold cross-validation results from the v1 sanity-check run (30 epochs/fold, Original Dataset, no augmentation, no group-aware splitting). Kept as a baseline for comparison with the headline v2 results.
+
+| Model | CV Acc (mean ± std) |
+|---|---|
+| VGG19              | 79.31 ± 1.89% |
+| ResNet101          | 79.27 ± 1.07% |
+| CLIP Transformer   | 63.92 ± 1.79% |