Investigating ML frameworks and libraries
The list of frameworks and libraries that help bring ML to edge devices is large and constantly evolving. This section will highlight the top three most common ML frameworks and libraries for Cortex-M ML at the time of writing this book. We will cover the following:
- TensorFlow Lite for Microcontrollers
- CMSIS-NN
- TinyML
TensorFlow Lite for Microcontrollers
The first framework to discuss is TensorFlow Lite for Microcontrollers (TFlite Micro). TensorFlow Lite is a framework that is designed for mobile devices, microcontrollers, and other edge devices where a small memory footprint and optimized performance are important. TFlite Micro is a C++ framework for ML inference that works well on Cortex-M microcontrollers. It is an optimized version of TensorFlow for embedded C/C++ applications. It is designed to run on 32-bit microcontrollers and results in very small binary files. It also doesn’t require an operating system...
