AI Projects with Raspberry Pi — out now!
Artificial intelligence (AI) isn’t just a data centre humming away inside a gigantic warehouse. It is also the single-board computer sitting on your desk, wired up to your smart home equipment, running models locally, securely, and privately.
This is the spirit behind our latest book: AI Projects with Raspberry Pi. This is a hands-on guide to discovering real-world AI applications with Raspberry Pi hardware. These aren’t toy demos, but rather projects you can build, run, and point to and say, “I made that”. Program it right and you might even get a spoken reply.
What’s inside
The book covers a broad range of AI applications, all of them running on a Raspberry Pi:
- Computer vision — recognise and classify what’s in front of a camera or inside files. We cover both image recognition and video recognition.
- Python and predictive modelling — work with the regression-style workhorses behind most practical data science.
- Speech to text (and vice versa) — convert speech to text, generate speech from text, and use spoken commands to control local devices.
- Linguistics — perform audio transcription, language translation, and processing.
- Sensor data — gather sensor data on a Raspberry Pi Pico. Use Python to train a machine learning model on your Raspberry Pi, then deploy the model on the Raspberry Pi Pico, where you’ll perform real-time inference.
- Large language models — run genuine LLMs locally on Raspberry Pi hardware and the Raspberry Pi AI HAT+ 2 accelerator. Run chatbot-style services securely, privately, and locally at the edge.
- Image generation — use Stable Diffusion techniques to generate images with Raspberry Pi hardware.
- Model training — discover how models are trained with test and training data. Slice and dice MobileNet and retrain it to detect hand gestures.
Raspberry Pi hardware… accelerated
If you are into AI and Raspberry Pi, this is the book for you. We make the most of Raspberry Pi hardware, both the bare board and systems accelerated with AI-specific hardware. Here, you’ll find projects that run on a range of Raspberry Pi devices, including Raspberry Pi 4, 5, and Zero 2 W as well as Raspberry Pi Pico microcontrollers.
As you’d expect, we make full and detailed use of the Raspberry Pi AI Camera and the Raspberry Pi AI HAT+ and AI HAT+ 2 — all using the standard Raspberry Pi OS, so you know exactly what you’re dealing with.
Behind the scenes of LLMs and generative technology
Generative technologies are a thorny subject. This book isn’t here to tell you right from wrong, or insist you get aboard the generative AI train. It’s here to help you understand the deployment and real-world usage of these technologies.
Once you move beyond the chatbots, you’ll discover a wide range of generative pre-trained transformer models that can be deployed to your computer. You’ll understand the training parameters, sizes, and weights that make up these models, learn how lightweight models are deployed on edge devices to perform inferencing, and discover how LLMs interact with code.
Because Raspberry Pi is a physical computer with input/output, you can use LLMs to perform real-world tasks.
The physical computing advantage
When it comes to AI, Raspberry Pi is different from a laptop computer or an online chatbot. Raspberry Pi is a real-world device operating at the edge, meaning you can wire it up to the physical world: sensors in, actuators out.
Take voice recognition, for example. In the consumer space, you speak into a device that sends your voice off to another computer in another part of the world, where it is analysed and a response is generated and sent back. For the end user, this is a black-box process — but in this book, you’ll come to understand speech to text and linguistics, and learn how to deploy the models that analyse speech and provide voice feedback for your project. Of course, because this is Raspberry Pi, you can use this knowledge to control any and all devices in your home, on the factory floor, or elsewhere in the real world.
Build real-world projects
Strap an accelerometer to your hand, wave it around, and you’re generating a stream of sensor data. Label the moments where a particular gesture actually happened, feed both the readings and the labels into a training process, and you end up with a model that can recognise that gesture in real time.
In this book, you’ll write code that watches for the gesture and, the moment it fires, tells an attached device to do something. We walk you through this by showing you how to make a magic wand that sparkles when you flick it. It’s one of several projects that take AI out of the abstract and into something you can hold in your hand.
Your journey through the book
The book starts with the basic vocabulary of AI and machine learning, including tensors, the foundational data structure behind almost everything that follows.
From there, it steps back to map out the Raspberry Pi AI stack, covering both hardware and software, before moving into regression models.
Things get visual with a chapter on generating images from text prompts, followed by a deep dive into running an LLM directly on your Raspberry Pi, using both the CPU and the AI HAT+ 2 accelerator.
Next comes the magic wand chapter: training a model to recognise gestures and building a wand that sparkles on cue.
Voice gets its own extensive chapter — transcribing speech, responding to spoken commands, generating speech from text, translating between languages, and even building a voice-powered chatbot.
The book then zooms in on image recognition, teaching you how to recognise and classify objects in real time using the Raspberry Pi AI Camera or Raspberry Pi AI HAT+ (version 1 or 2).
Finally, we look at accessing and deploying models from model zoos. And we take apart the MobileNet vision model, splicing off the end and retraining it in Python to perform hand gesture recognition.
Why it’s worth your time
If you’re the sort of person who wants to understand AI rather than simply use it, this book is for you. You’ll learn how models are trained, what they can and can’t do, where they’re genuinely useful, and where they’re overhyped. Most importantly, you’ll discover all of this by building practical projects on hardware you already know and love.
AI Projects with Raspberry Pi is available now from Raspberry Pi Press. You can pick up your copy from our store for just £17.99 £8.99 (introductory price). It’s also available from online retailers such as Amazon US and UK, and from other booksellers who have exceptional taste in books. If you’re interested in an electronic version, we’ve got several ways you can get your hands on a PDF or ePUB.
After you add this new book to your shopping cart, be sure to check out the many other books we offer in our online shop.
No comments
Jump to the comment form
Facts Only
* The book covers AI applications running on Raspberry Pi hardware.
* Applications include computer vision (image/video recognition), Python and predictive modeling, speech to text and vice versa, linguistics, sensor data processing, Large Language Models (LLMs), image generation using Stable Diffusion, and model training (e.g., MobileNet retraining).
* Projects utilize Raspberry Pi hardware, including the bare board, Raspberry Pi 4, 5, Zero 2 W, and Raspberry Pi Pico microcontrollers.
* The book details using the Raspberry Pi AI Camera and AI HAT+ / AI HAT+ 2.
* Model training involves methods like slicing and retraining MobileNet for hand gesture detection.
* LLMs can run locally on Raspberry Pi hardware and the AI HAT+ 2 accelerator.
* Sensor data can be gathered on a Raspberry Pi Pico, with models deployed for real-time inference.
* The book covers speech processing, including transcription, generation, translation, and voice-controlled device commands.
* The physical computing advantage involves wiring the Pi to sensors (input) and actuators (output).
* Project examples include training gesture recognition models and building interactive devices like a "magic wand."
Executive Summary
A book titled AI Projects with Raspberry Pi provides a hands-on guide to building real-world Artificial Intelligence applications using Raspberry Pi hardware. The content covers a broad spectrum of topics, including computer vision, predictive modeling using Python, speech-to-text and text-to-speech functionalities, linguistics, sensor data processing, running Large Language Models (LLMs) locally, image generation via Stable Diffusion, and model training techniques like hand gesture recognition for MobileNet. The guide emphasizes utilizing the Raspberry Pi hardware, including AI-specific accelerators like the AI HAT+ and AI HAT+ 2, alongside various models such as regression and transformer models.
The material positions the Raspberry Pi as a platform for deploying AI locally and privately, emphasizing an "edge computing" advantage where physical input/output devices connect to ML models. The book details a journey starting with foundational concepts like tensors, moving through the AI stack, and concluding with practical projects involving real-time inference and physical interaction, such as gesture recognition systems controlled by actuators.
Full Take
The narrative constructs an argument that positions edge computing, specifically via Raspberry Pi, as the necessary platform for genuine, embodied AI experience rather than abstract cloud-based services. The structure moves from theoretical foundations (tensors) to hardware integration and finally to complex deployment scenarios (LLMs, gesture recognition). This progression suggests a pattern of legitimizing localized, physical interaction as a superior mode of AI engagement over centralized systems.
The emphasis on deploying LLMs locally leverages the physical computer advantage: transforming abstract language models into tools that interact directly with the physical world via sensors and actuators. This implicitly frames local deployment not just as privacy protection, but as functional necessity—the ability to control real-world devices based on immediate environmental feedback. The focus shifts from *what* an AI can do (processing) to *where* it can act (physical execution).
The presentation of generative technologies is deliberately softened; the text avoids prescriptive stances on the ethics of generative AI, focusing instead on the mechanics of deployment and inference. This framing allows the work to invite technical exploration without imposing immediate ethical or philosophical debates upfront. The pattern suggests a strategic effort to build a community of practitioners by providing tools for tangible creation, where the learning pathway is intrinsically tied to physical manifestation—building systems that "do" things rather than just reporting data.
Bridge questions: How does this localized deployment paradigm shift the established relationship between end-users and monolithic AI services? What are the long-term implications of training highly customized, on-edge models versus continuously updated, centralized foundation models for real-world utility? If the physical interaction is paramount, what standards must evolve to govern the security and agency provided by these edge-deployed systems?
Sentinel — Human
The text reads like promotional material for a technical book, successfully weaving complex AI concepts with specific hardware applications, suggesting a strong human editorial voice guiding the narrative.