About this ebook
The book is aimed at readers who want to learn gradually and practically how to bring Artificial Intelligence to low-resource devices. With simple language, plenty of concrete examples, and step-by-step exercises, it provides the tools to move from theory to the creation of real-world applications.
Main Topics
The course develops through progressive lessons:
Introduction to the TensorFlow ecosystem for TinyML.
Conversion from TensorFlow to TensorFlow Lite to optimize models.
Transfer Learning and reuse of pre-trained models.
Techniques for optimization, quantization, and pruning.
Development of concrete applications:
Keyword Spotting (KWS) through voice recognition.
Visual Wake Words (VWW) for smart systems.
Anomaly Detection with K-Means, t-SNE, and Autoencoder/VAE.
Introduction to Knowledge Distillation.
Method and Practical Projects
Each concept is supported by Colab exercises with line-by-line comments, featuring replicable examples such as:
Training and deploying CNNs on microcontrollers.
Using MobileNetV2 in classification projects.
Building datasets with Pixabay API and Roboflow.
Implementing anomaly detection systems on real and synthetic data.
Key Strengths
Hands-on, progressive “lab-style” approach.
Complete workflow: from data to model with real optimizations.
Focus on memory, latency, accuracy, and responsible data usage.
Replicable applications in vision, audio, and anomaly detection.
Target Audience
Motivated beginners who want to bring AI to embedded systems.
Makers, IoT and Edge AI developers, data science students.
Prerequisites: basic knowledge of Python/Colab and Part 1 of this course.
Learning Outcomes
By the end of the book, you will be able to design and implement a complete TinyML app: collect and prepare data, train, optimize, and deploy models on microcontrollers. You will also gain access to reusable patterns for KWS, VWW, and Anomaly Detection, readily adaptable to new projects.
👉 In short: a practical, guided journey that takes you from idea to the real-world implementation of TinyML applications, opening the door to the world of Edge AI.
