Is it "bad" to load TensorFlow model via Arduino IDE instead of Ambiq SDK?

Development Boards Artemis
1

As I learn TensorFlow and machine learning in general, a tutorial I’m working up may benefit from simplicity where possible, focusing a bit more on the machine learning and less on new SDKs. I’m finding here that I can compile a model for Arduino IDE use (cool and great as I’m familiar with that):

https://github.com/tensorflow/tensorflo … ng-started

If I do this though, will I lose any benefit potentially gained if using the Ambiq SDK? For example, will it not be run from some special processing component that makes it faster, missed out if compiling for Aduino IDE? (sorry, still learning all this stuff)

Thanks for tolerating my silly question in advance!

2

Hi veyor,

I’m afraid I don’t have a great answer for you since the TensorFlow platform is not developed by SparkFun and you’re looking at an experimental branch here. From a quick look, you might be missing out on some features and might have some stability issues going with Arduino since it is still experimental.

If you have not already, the [TensorFlow Lite for Microcontrollers Guide and the other resources on that site will be a good place to learn more about using TensorFlow with your Artemis. [This tutorials page for the TensorFlow Core will also possibly have some good information for you.](Tutorials  |  TensorFlow Core)](TensorFlow Lite for Microcontrollers)

3

Hello TS-Mark,

Thank you for your response. As I dig deeper into C/C++, the fog is beginning to clear. I’m also seeing very recent Arduino-related TensorFlow Lite updates in the TensorFlow repository which helps as well. I may start from there an “level up” at that point if needed.

I’m having fun with the Qwiic-connected hardware!

Kind regards,

Chris