I’ve been using the Arduino Cloud for a little while, and I noticed the integration of Edge Impulse (great news). But then I realised that it’s restricted to the same means of data capture as Edge Impulse (I’ll call it Vanilla Edge Impulse).
I have numerous microcontrollers with a variety of different sensors. The two most basic are ESP32-based M5Stack Atoms. I have configured these to push their sensor data up to the Arduino Cloud directly as IoT things/devices via Wi-Fi without the need for a forwarder after their initial flash. And I get a nice dashboard, charts, etc. Basically, the variables whir away in the background gathering data, nice.
So, I’ve got my things/devices tucked into my Arduino Cloud, what I seem to be missing is that button either in Vanilla Edge Impulse or Arduino Edge Impulse to collect from Arduino Cloud.
Getting really creative, Arduino Cloud has emerging OTA support. Imagine if impulses could be pushed back OTA to my things/devices. Thus establishing a continuous loop from the edge and back again.
I’m thrilled you’ve brought up this topic, as I am currently working on a guide on how users and companies can effectively bridge the gap between standalone devices, their platforms, and Edge Impulse.
Your insights and use case are timely and align perfectly with our efforts to detail some of the current integrations, and an example of how to implement one on a stand alone platform like ESP32.
Our upcoming documentation will focus on several key aspects, and you are exactly right on the terminology we will be using:
Continuous Learning in Edge Impulse: Edge Impulse recognizes that machine learning models (impulses) are dynamic and require continuous updates as data evolves. This involves integrating OTA updates and continuous learning into the machine learning lifecycle. Partners are already integrating these aspects into their platforms. We will be gathering content on how these integrations were created so other platforms can follow.
Stay tuned for its release, and please don’t hesitate to reach out with any specific questions or insights you might have in the meantime! Your contributions are invaluable in shaping a practical and user-friendly resource for this.
Please share as much content as you can on the Arduino Cloud process you are using and I will try to include some details in the doc.