Running TensorFlow Lite on a device that is not officially supported

Projects General Guidance
1

Hello,

I am currently working on a project based on TensorFlow Lite. In the official TFLite documentation, there is a list with supported devices, which contains about 10 different MCUs including the Nano 33 BLE Sense. The docs also say that TFLite needs a 32-bit MCU.

So my question is: Does anyone know if it is possible to get TFLite models to run on other MCUs (which are not officially supported) that fulfill this requirement? Or is there some sort of device-specific library needed to run TFLite, so that it won't work on a device like, let's say, the Arduino Nano 33 IoT?

Thanks in advance!

2

What was the Tensor-Flow people's answer to your question?

3

More links please !
A link to the TensorFlow Lite, to that documentation, to that list, to the code for the Arduino, and so on.

The code is not easy: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/micro/examples.
It is not a simple sketch.
To adapt it for something else, everything about the code has to be known. That is a hard job.
I think they also use the sensors on the Nano 33 BLE Sense board.

This is interesting, but that does not seem to make it easier: TensorFlow Lite Ported to Arduino - Hackster.io.

4

This topic was automatically closed 120 days after the last reply. New replies are no longer allowed.