Introduction

Microcontrollers (MCUs) are cheap electronic components, usually with just a few kilobytes of RAM, and designed to consume small amounts of power. Today, MCUs can be found embedded in all residential, medical, automotive, and industrial devices. Over 40 billion microcontrollers are estimated to be marketed annually, and hundreds of billions are currently in service. But, curiously, these devices receive little attention because, many times, they are used just to replace functionalities that older electromechanical systems face in cars, washing machines, or remote controls.

More recently, with the era of IoT (Internet of Things), a significant part of these MCUs is generating “quintillions” of data, which, in their majority, are not used due to the high cost and complexity of their data transmission (bandwidth and latency).

On the other hand, in the last decades, we have witnessed the development of Machine Learning models (sub-area of Artificial Intelligence) trained with “tons” of data and powerful mainframes. But now, it suddenly becomes possible for “noisy” and complex signals, such as images, audio, or accelerometers, to extract meaning from the same through neural networks. More importantly, we can execute these models of neural networks in microcontrollers and sensors using very little energy and extract much more meaning from the data generated by these sensors, which we are currently ignoring.

TinyML, a new area of applied AI, allows the extracting of “machine intelligence” from the physical world (where the data is generated).

TinyML Made Ease is a foundational text designed to facilitate the understanding and application of Embedded Machine Learning, or TinyML. In an era where embedded devices boast ultra-low power consumption—measured in mere milliwatts—and machine learning frameworks such as TensorFlow Lite for Microcontrollers (TF Lite Micro) become increasingly tailored for embedded applications, the intersection of AI and IoT is rapidly expanding. This book demystifies the integration of AI capabilities into these devices, paving the way for a wide-reaching adoption of AI-enabled IoT, or “AioT.”