Line following example training

[aitorarrietamarcos]

Hi,

We are trying to run the line following example training. However, for the current training setup in the notebooks, the algorithm stops training by the 22nd epoch. We have tried different things like increasing the batch size, but it stops the training before. The obtained models are not working as the pre-trained models work.

Do you have any hint on where the problems can be?

[Bitterisland6]

Hi,
I'm almost sure that your problem is caused by EarlyStopping callback from keras. You can remove, or comment it and try to run the training again (callbacks are defined one cell above the cell with training).

You can also try to change the parameters that are specified in the callback.I wasn't experienced with the EarlyStopping when i was making the notebook, so there is a chance something there may be done better (i guess changing monitor and min_delta parameters to something more precise will help)

[Bitterisland6]

You can also try to run the training loop on the same model a few times - after one training finishes, don't run the cell with model definition. Instead run again the training on the same model - this is not a pretty solution, but it will work.

And sorry if my suggestions are obvious to you, I don't know how experienced with neural networks you are. 😅