Figure 2.

Training curves showing decrease in generalized dice loss for the training (red “+”) and validation (green “x”) as training progressed. The model with the lowest validation loss was chosen for the final model, to reduce overfitting.