Pre-trained network on the raspberry pi

Hello everyone,

I would like to train a Deep Learning Network in Matlab to recognize coughs from a background sound. Then I want to use this network on a Raspberry Pi 4B. I don't care if it runs on the Raspberry as a Matlab executable or in an other language like python.

I want the pi to work in such a way that:

• Starts recording a 10 sec audio clip using a Python script and saves it on the pi.
• Then the clip is cut into a 1 second clips and its mel-spectrogram images are formed saved on the pi
• Then using MATLAB the pi takes the sample and uses a pre-trained network for classification and outputs the result.

I wan to know if it can all be done independently. I have a modivius( intel neural stick 2 for hardware accceleration used for Deep learning)

I also trained a network on my laptop, but only use it for classification on that same laptop after.

I am a little confused now, what I have to include in the code running on the Raspberry. I want to used the trained network, so probably not the train-functions?

How and where exactly is the trained network stored and how do I get access to it? Could I also use the in matlab trained network in Python?

Would be great if someone can help me.

Answer 1

[jibrahim]

Hi Joenam,

This example might be helpful to you:

Speech Command Recognition Code Generation on Raspberry Pi