TinyML Cookbook: Combine artificial intelligence and ultra-low-power embedded devices to make the world smarter

I found this book really well written, with a refreshing clarity of wording, nice accessible explanations of each new topic or terminology, and a range of fun and interesting example TinyML applications to learn from by doing.The book is nicely structured and well paced, giving everything from the context, purpose, software instructions and links, and even the hardware shopping list with online links to help get everything needed to try out the examples.The chapters go from very first introduction through to giving several different fully explained and easy to follow examples to implement TinyML systems for yourself.Chapter list:1. Getting Started with TinyML2. Prototyping with Microcontrollers3. Building a Weather Station with TensorFlow Lite for Microcontrollers4. Voice Controlling LEDs with Edge Impulse5. Indoor Scene Classification with TensorFlow Lite for Microcontrollers and the Arduino Nano6. Building a Gesture-Based Interface for YouTube Playback7. Running a Tiny CIFAR-10 Model on a Virtual Platform with the Zephyr OS8. Toward the Next TinyML Generation with microNPUMost of the examples show how to run on one of the following micro-controller boards:* Arduino Nano 33 BLE Sense* Raspberry Pi PicoThere are also a couple of examples running on a virtual platform instead of physical hardware.(Somewhat inevitably for any non-trivial technical exercises, I did spot a few minor errata. The publisher does helpfully provide a feedback form, and has said the errata will be published on the GitHub to help other readers, so please do look for that or ask the publisher if you need to.)To run the examples on real hardware there are some electronic components required, including the micro-controller itself but then also solder-less breadboards, sensors, jumper wires, battery holders and a few basic components such as resistors or LEDs.The link to the shopping list on publisher’s GitHub page is given in the book, but if you want to save yourself some time and order at the same time as the book, I summarise the shopping list as follows:* Arduino Nano 33 BLE Sense with headers: https://www.amazon.co.uk/Arduino-NANO-SENSE-headers-mounted/dp/B07WXKDVTL* Raspberry Pi Pico with headers: https://www.amazon.co.uk/Raspberry-Soldered-Microcontroller-Development-Pre-soldered/dp/B08ZSMZ8FT* Electronics Component Basic Starter Kit: https://www.amazon.co.uk/gp/product/B01LXTH7U1* Half-size solder less breadboard: https://www.amazon.co.uk/gp/product/B0739XRX8F* Temperature and Humidity Sensor Module DHT22/AM2302: https://www.amazon.co.uk/Digital-Temperature-Humidity-Electronic-Practice/dp/B091CP63LH* Breadboard jumper wires; male-to-male: https://www.amazon.co.uk/gp/product/B01LXTI2E3* Breadboard jumper wires; male-to-female: https://www.amazon.co.uk/AZDelivery-MB-102-Breadboard-Kit/dp/B07K8PVKBP* Battery holders for 3xAA and 4xAA: https://www.amazon.co.uk/gp/product/B08Z3RJC1Y* Camera module OV7670: https://www.amazon.co.uk/gp/product/B0919PWH4Q* Accelerometer sensor module MPU-6050: https://www.amazon.co.uk/gp/product/B07P5YXBXVSome of the links above are for larger quantity than strictly needed, so you might want to look for other listings or other retailers. Amazon review rules only allow me to include links within their site, but there are many other relevant retailers too such as thepihut, arduino store, adafruit etc.This is a great book to demystify the terminology of ML, get a feel for what it is really about, and try it out for yourself. Have fun!

As someone who has created one of the the projects in the book from stratch I can definetly say this book makes the process a lot easier both with the code and guidance provided.Would recommend getting the nano 33 for someone new to hardware as it has most of the components needed for the examples already integrated.Only negative was the fact that the rp pico can isn't always included as an alternative to the nano 33 on some of the examples even though it is capable (though this might have been excluded due to the extra level of complexity in embedded coding required).