Browse tools

The basics

Computer vision

In short

Computer vision is the branch of AI that lets computers make sense of images and video: recognising objects, reading text, finding faces and tracking movement. It is behind photo search, plant ID apps, chatbots that read pictures and police facial recognition.

Also called
CV, Machine vision, Image recognition
Topic
The basics
Level
Beginner
First described
An early statement of the goal is Seymour Papert's MIT memo "The Summer Vision Project", July 19661.

A computer sees a photo as a grid of numbers, one per pixel. Computer vision is about turning those numbers into useful answers, such as "this is a cat" or "this face matches that one".

In July 1966 Seymour Papert at MIT set out a "Summer Vision Project" for summer workers, with the final goal of object identification: naming objects by matching them against a known vocabulary1.

Modern computer vision rests on two things. First, data: ImageNet, introduced in 2009, is a huge collection of labelled pictures organised by the WordNet hierarchy of word meanings23 and now indexes over 14 million images3. Second, deep learning: in 2012 Krizhevsky, Sutskever and Hinton trained a deep convolutional neural network on 1.3 million ImageNet images and cut error rates well below the previous best4.

Since then, vision has merged with language. Models such as OpenAI's CLIP learn from images paired with captions, so you can describe what to look for in plain words5.

Computer vision now raises legal questions in the UK. Live facial recognition turns faces into biometric values and compares them with a police watchlist6. In 2020 the Court of Appeal allowed Mr Bridges' challenge to South Wales Police's use of automated facial recognition on three grounds7. In April 2026 the High Court dismissed a challenge to the Metropolitan Police's live facial recognition policy6.

Example

Near London Bridge station, a man was stopped and questioned after the Metropolitan Police's live facial recognition system matched him with his brother, who was on a watchlist. The case reached the High Court in 20266.

Why it matters

Cameras that can recognise what they see are now in phones, shops, cars and police vans. That brings convenience, but also mistakes and privacy questions that UK courts are still working through.

Questions people ask

What is computer vision used for?

Recognising objects and faces, reading text in photos, sorting your photo library, identifying plants, guiding robots and cars, and spotting people on police watchlists6. Chatbots that let you upload a photo also use it5.

Is facial recognition legal in the UK?

Police use is allowed, but it must meet legal requirements6. In 2020 the Court of Appeal ruled against South Wales Police on three grounds7. In 2026 the High Court found the Metropolitan Police's live facial recognition policy lawful and dismissed a challenge to it6.

How does computer vision work?

Most systems today use deep neural networks trained on millions of labelled images. The influential 2012 network by Krizhevsky and colleagues had 60 million parameters and five convolutional layers4.

Related terms

Convolutional neural network Deep learning Vision-language model Multimodal AI AI and privacy Deepfake Training data

Tools that use it

Sources

  1. Papert, S. A. (1966), The Summer Vision Project, MIT AI Memo AIM-100
  2. Deng et al. (2009), ImageNet: A large-scale hierarchical image database, IEEE CVPR 2009
  3. ImageNet, About ImageNet
  4. Krizhevsky, Sutskever and Hinton (2012), ImageNet Classification with Deep Convolutional Neural Networks, NeurIPS 2012
  5. Radford et al. (2021), Learning Transferable Visual Models From Natural Language Supervision (CLIP), arXiv:2103.00020
  6. R (Thompson) v Commissioner of Police of the Metropolis [2026] EWHC 915 (Admin), paras 2, 17, 229-231
  7. R (Bridges) v Chief Constable of South Wales Police [2020] EWCA Civ 1058, para 209

Back to the glossary