What is Computer Vision (How Do Machines See)?

Computer vision is a branch of AI that helps machines "understand" images and video. But it doesn’t see like we do. It reads pixels, analyses patterns, and matches them to what it’s learned before.

Think of it like teaching a kid to recognise animals. You show them loads of pictures… “This is a dog, this is a cat, that’s a bird.” As time goes on, they start to guess the animal correctly, even if the image that you’re showing them isn’t exactly the same as the one before.

The representation for computer vision is the same, except it’s trained on millions of images, not just a handful.

The AI learns to map visual features to labels. It might learn that pointy ears, whiskers, and certain eye shapes usually mean "cat." When it sees a new image, it compares the visual features to patterns it learned during training.

How the Technology Actually Works

You might be left wondering “How does it actually work?”

Well, most modern computer vision uses deep learning, specifically something called convolutional neural networks (CNNs).

Here’s the short version:

1. It starts with raw pixels, just numbers representing colours and brightness
2. The AI breaks the image into tiny pieces and processes them layer by layer
3. It first detects simple stuff like edges and shapes
4. Then it combines those into more complex features, like “ears,” “eyes,” “text,” or “car”
5. Finally, it puts it all together and assumes the image