Model Training early stopping about autonomousdrivingcookbook

[mitchellspryn]

This is expected and desired. Notice how your validation loss is not improving - this means that if you continue to train, your model will overfit. You can see the effects of that by removing the EarlyStoppingCallback from your code and retraining - you will most likely get poor results.

How does the trained model perform? Does it seem to predict the correct angles (in the last cell of the notebook)? How far does the car go before crashing? Note that since we don't have a very large dataset, it's possible that the model converges to a suboptimal minimum. Retraining will help in that case.