Update README.md

README.md

@@ -79,6 +79,10 @@ The model uses a sliding window approach to capture temporal patterns in musical
 - **MIDI-Only**: Limited to MIDI format; cannot process audio recordings or sheet music
 - **Cultural Bias**: Training data may reflect Western classical music traditions
 
+Below is the confusion matrix for the best-performing checkpoint, visually highlighting these misclassifications (click to enlarge):
+
+[<img src="confusion_matrix_best.png" alt="Confusion Matrix" width="500"/>](confusion_matrix_best.png)
+
 ### Recommendations for Use
 - Validate results with musicological expertise, especially for Classical period identification
 - Use confidence thresholds to filter low-confidence predictions
@@ -132,6 +136,7 @@ The following hyperparameters were used during training:
 
 ### Training results
 
+
 | Training Loss | Epoch  | Step  | Validation Loss | Accuracy | F1     |
 |:-------------:|:------:|:-----:|:---------------:|:--------:|:------:|
 | 1.2797        | 0.1031 | 2000  | 1.3522          | 0.4608   | 0.2486 |
@@ -184,8 +189,14 @@ The following hyperparameters were used during training:
 | 0.8264        | 4.9508 | 96000 | 1.1056          | 0.5736   | 0.4174 |
 
 ### Training Analysis
+
+Below is the full training metrics plot, showing loss, accuracy, and F1-score trends over the entire training process (click to enlarge):
+
+[<img src="training_metrics.png" alt="Training Metrics" width="500"/>](training_metrics.png)
+
 The training shows stable convergence with the model reaching its best performance around step 44,000 (epoch 2.27). The training loss decreases steadily while validation metrics stabilize, indicating good generalization without severe overfitting. The model achieves its peak F1 score of 0.4299 at step 44,000, which was selected as the best checkpoint.
 
+
 ### Framework versions
 
 - Transformers 4.49.0