Image classification in openmv m7

Hi, I am new to machine learning and using Open MV in general. I have purchased the Open MV M7, and want to do object classification with it in real time. Is this possible on this model? If so, what would be the maximum size of the .tflite model that we could use?

Hi, it is possible, however, memory is very limited. You should make a network that has no larger than a 48x48 or 32x32 input network size.

Hi, thanks for the prompt reply.

Can we use Edge Impulse to create such a model? Or is this only possible with Open MV H7 plus? I remember reading in a blog post that edge impulse only supports the H7.

If it is not possible to use edge impulse, are there any other alternative APIs or softwares that we can use to achieve that goal?

Yeah, you can do this via Edge Impulse. Just select the the Image input to not be 96x96 by default. There’s an option when you are building the network.

Could we also just use transfer learning and retrain a mobileNet model? Or would this not fit into our M7 module?

That’s literally what Edge Impulse is doing. Just use their setup.