Training an edge–cloud-based model with Edge Impulse

I am trying to design a model using online machine learning method. Basically, I want the training data generated by Raspberry Pi 4 model B to be sent to the Edge Impulse platform, which would then perform the training and finally deploy the trained model back to the raspberry Pi. Since my data will be generated as a function of time and I want the model to dynamically adapt to new patterns in the data, I want the model to be updated at intervals when a new data arrives and after the training, the newly trained model should be deployed back to the Raspberry Pi. I want the entire process to be automated. I am wondering whether this is possible with edge impulse or any other platform. I am open to any suggestions that can be helpful.
Any help will be much appreciated.
Thanks.

Hi @Vincent,

Yes this is possible, there are many ways you can send the training data from your Raspberry Pi to Edge Impulse, here’s just a few of the options:

Then, after you have uploaded all of your data to your training and testing sets of your Edge Impulse project, and trained your machine learning model, then you can deploy your model down to your Raspberry Pi and start inferencing on new, un-seen data with the edge-impulse-linux-runner: https://docs.edgeimpulse.com/docs/raspberry-pi-4#deploying-back-to-device

And of course, all of these options can be automated via a Python script, shell script, etc.

Please let me know if you have any questions!

@jenny Thanks for the prompt response.
I will check the links you sent and will keep you updated.
Do you also have any link that can shade light on how to automate the whole process? Because that is key for me, I want the model to be retrained with new data and re-deployed to the raspberry pi automatically when the new data fulfills certain conditions.
Thanks.