YOLOv5 bring your own model retraining

Help
#1

Hi everyone

I followed the YOLOv5 transfer learning model for Edge Impulse tutorial and every step works as expected. Thanks for writing this up in detail!

Now that I would like to use the model in the Edge Impulse GUI, I got stuck on the retraining of the model. Let’s say that I’m training the model extensively on my build machine and then pushing it back to EI. Is there a possibility to use the model as I pushed it in the ‘Object detection’ step?

I’d like to set up the following flow:

  • Add images to EI and annotate them
  • Download new data to build machine
  • Retrain model extensively
  • Push model to EI
  • Use model as is

For now, the only option to proceed is:

  • Add images to EI and annotate them
  • Download new data to build machine
  • Retrain model extensively
  • Push model to EI
  • Retrain model on EI to be able to proceed to next steps

Thanks for looking into this.
Kind regards!

#2

Hello @aTEraGOU,

FYI, we just released a new version of BYOM: https://www.edgeimpulse.com/blog/unveiling-the-new-edge-impulse-python-sdk
This will let you import directly your model.

Best,

Louis