Loss Curves

The loss measures how different the model's predictions are from your annotations — lower is better.

Loss is a numerical measure of prediction error. The chart shows two curves:

Training Loss (red): Error on the patches used for learning. Should decrease steadily as the model learns.

Validation Loss (blue): Error on held-out patches not used for training. Shows how well the model generalizes to unseen data.

What to look for

Healthy Training: Both curves decrease together and eventually flatten. The validation loss may be slightly higher than training loss.

Overfitting: Training loss continues to decrease but validation loss starts increasing. The model is memorizing training data rather than learning generalizable patterns. Consider reducing epochs or adding more training data.

Underfitting: Both losses remain high and don't decrease much. The model isn't learning effectively. Try increasing model capacity (more features/layers) or training longer.

Instability: Losses jump around erratically. Try reducing the learning rate for smoother optimization.