I went through the training progress plots you have shared. It doesn’t look like your model is overfitting. Here's how I am interpreting the different lines in the plot and why I believe that your model is generalized enough:
1. Accuracy Plot (Top plot)
- The blue solid line represents training accuracy.
- The black dashed line with dots represents validation accuracy.
These two lines track very closely, and both plateau at high values (training approximately 99%, validation at 98.08%). The fact that validation accuracy stays high and doesn't lag significantly behind training accuracy is a good sign and it means your model is learning useful patterns and not just memorizing the training data.
2. Loss Plot (Bottom)
- The orange solid line represents the training loss.
- The light orange dashed line represents the validation loss.
Since both the losses decrease smoothly and stay low throughout the training process and the validation loss never increases while training loss continues to drop (increasing validation loss while decreasing training loss is a classic symptom of overfitting).
