Accelerometer data

My accuracy of my model suddenly drops when I run the model test option and I am not able to understand what I am missing. I have very clean separation in the features but I assume my data collection is clean.

1 Like

Hi @abhatikar,

Could you kindly share your project ID?

Thanks,
Clinton

https://studio.edgeimpulse.com/studio/82177

1 Like

Any help here ? @artiebeavis

I am sorted, thank you. My model was overfitted. :smiley:

1 Like

Hi @abhatikar

Nice to hear you found a solution. You could try adding some more data and retrain Thanks,
Clinton

1 Like

Hello @abhatikar,

Indeed you can see in the confusion matrix that only one class was detected properly.
To reduce overfitting here are a few things you can quickly do:

  • Add more data
  • Add dropout layers
  • Reduce learning rate
  • Add a more complex NN architecture

Regards,

Louis

1 Like