What does a computer “see” in a picture?
Computer vision treats images and video as data in which it looks for objects, patterns or movement.
What is it?
Computer vision treats images and video as data from which a model recognises objects, faces, text or movement. A computer does not see as a person does: it computes with pixel values and patterns it has learned.
An everyday example
A car park reads your number plate, and your phone recognises text in a photo so you can copy it.
Why does it matter?
Vision is used in health care, transport, industry and security. Camera surveillance also raises privacy questions.
A common misconception
That a model which recognises an object has understood the scene. A small change to the image or unusual lighting can fool it.
In more detail
What can it recognise?
It can detect a face, a vehicle, a tumour in a scan or a change in a satellite image — given a suitable task and proper testing.
Does it see like a human?
No. To the system, an image is numerical input. Its interpretation can fail with different lighting, angles or surroundings.
Is detecting the same as understanding?
No. Naming the objects does not necessarily mean grasping the whole context.
When is a wrong detection dangerous?
When it leads to decisions about people’s health, safety, surveillance or rights.
Remember: “It saw” is a simplified way of describing a calculation.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.