Book Image

TinyML Cookbook

By : Gian Marco Iodice
Book Image

TinyML Cookbook

By: Gian Marco Iodice

Overview of this book

This book explores TinyML, a fast-growing field at the unique intersection of machine learning and embedded systems to make AI ubiquitous with extremely low-powered devices such as microcontrollers. The TinyML Cookbook starts with a practical introduction to this multidisciplinary field to get you up to speed with some of the fundamentals for deploying intelligent applications on Arduino Nano 33 BLE Sense and Raspberry Pi Pico. As you progress, you’ll tackle various problems that you may encounter while prototyping microcontrollers, such as controlling the LED state with GPIO and a push-button, supplying power to microcontrollers with batteries, and more. Next, you’ll cover recipes relating to temperature, humidity, and the three “V” sensors (Voice, Vision, and Vibration) to gain the necessary skills to implement end-to-end smart applications in different scenarios. Later, you’ll learn best practices for building tiny models for memory-constrained microcontrollers. Finally, you’ll explore two of the most recent technologies, microTVM and microNPU that will help you step up your TinyML game. By the end of this book, you’ll be well-versed with best practices and machine learning frameworks to develop ML apps easily on microcontrollers and have a clear understanding of the key aspects to consider during the development phase.
Table of Contents (10 chapters)

Quantizing the model with the TFLite converter

Exporting the trained network as SavedModel saves the training graphs such as the network architecture, weights, training variables, and checkpoints. Therefore, the generated TF model is perfect for sharing or resuming a training session but not suitable for microcontroller deployment for the following reasons:

  • The weights are stored in floating-point format.
  • The model keeps information that's not required for the inference.

Since our target device has computational and memory constraints, it is crucial to transform the trained model into something compact.

This recipe will teach how to quantize and convert the trained model into a lightweight, memory-efficient, and easy-to-parse exporting format with TensorFlow Lite (TFLite). The generated model will then be converted to a C-byte array, suitable for microcontroller deployments.

The following Colab file (see the Quantizing the model with TFLite converter section...