Support for MAX78000?

Feature requests
#1

Hi team. There is no sign of the MAX78000 on your website. Do you have plans to support it?

The chip pairs a Cortex processor with a custom CNN block and claims spectacular power savings over software-only NN processing. But NN development is on Linux machines with GPUs.

There is a nice, cheap evaluation board: https://www.maximintegrated.com/en/products/microcontrollers/MAX78000FTHR.html

If you are not supporting it then I wonder if your tools could be used to make a model, and us users manipulate that with Maxim’s tools?

1 Like
#2

Hi @acutetech I’m not sure what the toolchain for the Maxim board is but you can probably:

  1. Export as C++ Library and use the code for your DSP blocks as-is.
  2. Take the ‘TensorFlow Lite (int8 quantized)’ model from Dashboard in your Edge Impulse project and convert it with the Maxim tools.
  3. Invoke the model from your application using the Maxim SDK.