3 ms·

Inferencing is the focus for now. We should able to go from TF to MCU (mbed) without writing any NN code.