As an IoT developer, you might think of machine learning as a server-side technology. In the traditional view, sensors on your device capture data and send it to the cloud, where Machine Learning (ML) models on hefty machines make sense of it. A network connection is obligatory, and you are going to expect some latency, not to mention hosting costs.
But more and more, developers want to deploy their ML models to the edge, on IoT devices themselves. If you bring ML closer to your sensors, you remove your reliance on a network connection, and you can achieve much lower latency without a round trip to the server.
This is especially exciting for IoT because less network utilization means lower power consumption. Also, you can better guarantee the security and privacy of your users because it does not require you to send data back to the cloud unless you know for sure that it is relevant.
In the following guide, you will learn how you can perform machine learning inference on an Arm Cortex-M microcontroller with TensorFlow Lite for Microcontrollers.
About TensorFlow Lite
TensorFlow Lite is a set of tools for running machine learning models on-device. TensorFlow Lite powers billions of mobile app installs, including Google Photos, Gmail, and devices made by Nest and Google Home.
With the launch of TensorFlow Lite for Microcontrollers, developers can run machine learning inference on extremely low-powered devices, like the Cortex-M microcontroller series. Watch the following video to learn more about the announcement: