Last checked: 6 November 2026
In short: Computer vision is the field of AI that lets computers interpret images and video. It can spot objects, read text, recognise faces and check products for faults, often faster than a person watching a screen.
Table of Contents
How it works
- Capture. A camera or scanner turns a scene into pixels, the tiny coloured dots that make up a digital image.
- Learn patterns. A model trained on millions of labelled images learns which pixel patterns mean a cat, a crack or a number plate.
- Decide. The system names what it sees, marks where it is in the picture and can trigger an action, such as a warning.
Where you’ll meet it
Face ID on an iPhone, Google Lens, number-plate cameras in car parks and photo search in Google Photos all use computer vision.
What it means for your business
- Cameras can do inspection work. Factories, farms and warehouses use it to count stock, check quality and flag safety problems.
- Labelled images cost money. You may need thousands of tagged photos of your own products before a model works well.
- Edge cases decide the payback. Bad lighting and crowded scenes can force people to check results, which eats into the savings.
Watch out for
- Privacy law. Filming staff or customers, especially with face recognition, brings strict data protection duties in the UK and EU.
- Hidden human labour. Some “automatic” systems rely on people reviewing footage behind the scenes.
Example
Amazon. In April 2024 Amazon said it would remove Just Walk Out, its camera and sensor checkout system, from its US Amazon Fresh grocery stores. It switched to Dash Carts, smart trolleys that identify items as shoppers add them. Amazon said customers wanted to see receipts and savings while they shopped. Reports said many transactions had needed human review, which Amazon disputed.
Computer vision vs multimodal AI
| Computer vision | Multimodal AI | |
|---|---|---|
| What it handles | Images and video only | Text, images, sound and video together |
| Typical job | Spot, count or check things in a picture | Answer questions about a mix of inputs |
| How they relate | One sense on its own | Combines vision with language and other senses |
Frequently asked questions
Is computer vision the same as face recognition?
No. Face recognition is one use of computer vision. The same methods also read text, count stock, guide robots, check medical scans and spot faults on production lines.
Can a small firm afford computer vision?
Often, yes. Off-the-shelf tools and cloud services charge per image or per camera, so a simple stock count or quality check can start for a modest monthly fee.
How much data does a computer vision project need?
It depends on the task. Common objects need few of your own images because ready-made models already know them, while rare faults may need thousands of labelled examples.