NN Block, validation and training loss graphs

[nash]

I want to see what my training and validation graphs look like. Is there a way to modify the NN block so that I can visualize my validation and training losses?

Thank you in advance.

[louis]

Hello @nash,

Feel free to have a look at this discussion which cover the same topic:

Plot loss and accuracy graph for object detection

Hi @yahyatawil @yodaimpulse actually we’ll probably not add it (and the reason why we haven’t added it in the past) because I think it’s bad UX and should be something the computer can figure out much better. Things that you can deduce from this graph, like ‘is my model still learning’, ‘is my model overfit’ should be an integral part of the training pipeline and be managed for the developer. Rather than showing a graph and calling it solved we’re implementing early stopping, automatically indic…

You can also retrieve the logs from the API if you want to do something with it:

Getting job ID: https://docs.edgeimpulse.com/reference/edge-impulse-api/jobs/list-finished-jobs
Getting job logs: https://docs.edgeimpulse.com/reference/edge-impulse-api/jobs/get-logs

Also, you can always export a Jupyter notebook and use matplotlib or TensorBoard directly.

Best regards,

Louis