Enable all 5 images: training curves, confusion matrix, per-class accuracy, confidence histogram, misclassification examples

README.md

@@ -160,31 +160,11 @@ Four candidate models were evaluated on the validation set:
 | 5 | 95.83% | 0.4130 |
 | **SWA Average (final)** | **95.93%** | **0.3981** |
 
-<!-- IMAGE PLACEHOLDER: Training curves (accuracy + loss across all phases)
-     Upload the training curves image from Cell 39 of the V15 notebook.
-     File to upload: cell39_accuracy_loss_curves.png (~68 KB)
-     This shows Phase 1 (1 epoch, val_acc 86.71%), Phase 2 (6 epochs, val_acc 93.35%),
-     and SWA continuation (5 epochs to val_acc 95.93%).
-     Uncomment the line below after uploading:
 ![Training Curves](training_curves.png)
--->
-
-<!-- IMAGE PLACEHOLDER: Confusion matrix heatmap
-     Upload the confusion matrix image from Cell 35 of the V15 notebook.
-     File to upload: cell35_confusion_matrix.png (~51 KB)
-     Shows percentage per row with raw counts, highlighting confusion between
-     temple↔mosque and stadium↔temple pairs.
-     Uncomment the line below after uploading:
+
 ![Confusion Matrix](confusion_matrix.png)
--->
 
-<!-- IMAGE PLACEHOLDER: Per-class accuracy bar chart
-     Upload the per-class accuracy image from Cell 40 of the V15 notebook.
-     File to upload: cell40_per_class_accuracy.png (~39 KB)
-     Shows stadium as the lowest class and skyscraper as the highest.
-     Uncomment the line below after uploading:
 ![Per-Class Accuracy](per_class_accuracy.png)
--->
 
 ## Training Details
 
@@ -464,13 +444,7 @@ print(f"Predicted: {LABELS[np.argmax(preds)]} ({np.max(preds)*100:.1f}%)")
 - **Skyscraper↔Bridge** confusion occurs when bridge structural elements resemble tall buildings against skyline
 - Inference confidence can be low on atypical examples (e.g., skyscraper predicted at 57.0% confidence with bridge at 36.2%)
 
-<!-- IMAGE PLACEHOLDER: Misclassification examples
-     Upload the misclassification grid from Cell 40 of the V15 notebook.
-     File to upload: cell40_misclassification_grid.png (~2.3 MB, resize recommended)
-     Shows up to 9 concrete failure cases with true vs predicted labels and confidence.
-     Uncomment the line below after uploading:
 ![Misclassification Examples](misclassification_examples.png)
--->
 
 ## Ethical Considerations