Computer Vision: How AI Sees
The big idea: a computer does not "see" a picture the way you do. To a computer, every image is just a giant grid of numbers. Computer vision is the branch of AI that turns those raw numbers into meaning — "this is a cat", "there is a pedestrian on the left", "this mole looks suspicious". It does this by building up understanding in layers: from tiny edges, to simple shapes, to object parts, to whole objects.
By the end of this lesson you will be able to:
- Explain how an image is stored as a grid of pixel numbers, and what RGB means.
- Describe the feature hierarchy: edges shapes parts objects.
- Tell apart the four main vision tasks: classification, detection, segmentation, and face recognition.
- Name real applications: phone unlocking, medical scans, self-driving cars, produce sorting, security.
- Explain where vision systems fail: unusual angles, poor light, adversarial stickers, and biased face datasets.
1. An Image Is a Grid of Numbers
Zoom into any digital photo far enough and it dissolves into tiny colored squares called pixels (short for "picture elements"). Each pixel stores a number — or a few numbers — describing its brightness and color. A modern phone photo has around 12 million pixels; that is 12 million little numbers (or 36 million, in color) that together form the picture.
In a grayscale (black-and-white) image, each pixel is a single number from 0 (pure black) to 255 (pure white). Middle values are shades of gray. Here is a tiny 6×6 grayscale image of a bright diagonal stroke on a dark background — first as numbers, then as shading: