Slides and materials for U of A CompSci AI club seminar
The Google slides Embedded Machine Learning
provide all the main details to train and deploy a speech recognition ("wake word") model to an microcontroller platfrom. The specific example is for the Artemis module/board.
The Jupyter notebook CompSciAITF115.ipynb
is lightly documented to walk through training a small model. Other model architectures can be tried to be sure, so whack away!
The speech data set is speech_commands
The Apache 2.0 license is used to be consistent with the code that has been copied from TensorFlow.org.