This book is licensed for the personal use of the buyer. Please do not share or resell the file. Short quotations with attribution are welcome.
The information in this book is provided for education. It is not legal, financial or other professional advice. Product and company names are used only to describe them and belong to their owners.
This is a free sample: the introduction and first chapter.
IntroductionIntroduction
Computer vision is the part of AI that makes sense of images and video. It unlocks phones with a glance, reads number plates, sorts parcels, inspects products on factory lines, helps doctors read scans, counts wildlife from drone footage and lets cars see the road. It is also one of the areas where AI's progress has been most visible: tasks that defeated computers for decades became routine in a few years.
This book explains how computer vision works and how to build vision systems that hold up in the real world. It starts with what an image is to a computer, moves through classical image processing and convolutional networks, and then covers the main tasks: classification, detection, segmentation, text recognition and more. Later chapters address vision transformers, image embeddings, multimodal models, data and labelling, evaluation, robustness, fairness, privacy and deployment. Two case studies show how the pieces come together.
Worked examples use small numbers you can check on paper. No advanced mathematics is required, though familiarity with basic neural networks helps.
How the book is organised
Chapters 1 to 12: images as data, classical processing, convolutions and network architectures.
Chapters 13 to 18: classification, data, labelling, augmentation, transfer learning and evaluation.
Chapters 19 to 27: detection, segmentation, keypoints and text recognition.
Chapters 28 to 32: vision transformers, embeddings, vision–language models, video and 3D.
Chapters 33 to 37: robustness, fairness, privacy, deployment and monitoring.
Chapters 38 to 40: medical and scientific imaging, satellite and aerial imagery, and generating and editing images.
Chapters 41 to 43: two case studies and a project checklist.
Chapter 1How Machines See
Humans recognise a cat instantly, whatever its colour, pose or lighting. For a computer, the same cat is a grid of millions of numbers that changes completely when the cat moves a few centimetres or the sun goes behind a cloud.
Why vision is hard
Viewpoint: the same object looks different from every angle.
Lighting: shadows and colour casts change pixel values dramatically.
Scale: objects appear large or tiny.
Occlusion: parts are hidden behind other things.
Background clutter: objects blend into busy scenes.
Variation within a class: chairs come in countless shapes.
The modern approach
Instead of writing rules for each case, modern vision learns from many labelled examples. Neural networks discover the features that matter, from edges up to whole objects, by adjusting millions of weights to reduce errors.
You have reached the end of the free sample. The full book continues with 42 more chapters, plus the appendices.