B4 · Domains of AI: Vision, Language, Speech and Robotics

Computer Vision Around Us

Computer vision is used wherever looking carefully, repeatedly and quickly is valuable: reading medical scans, checking crops, managing traffic, running shops, analysing sport and helping people who cannot see well.

6 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

How many cameras did you walk past today that were doing more than recording?

Computer vision is used wherever looking carefully, repeatedly and quickly is valuable: reading medical scans, checking crops, managing traffic, running shops, analysing sport and helping people who cannot see well.

In simple words

Anything a trained person does mainly with their eyes, a vision model can often help with: a doctor studying an X-ray, a farmer checking a leaf, an umpire watching a ball, a cashier scanning fruit. The model does the looking. A person usually still makes the decision.

Why it matters

Cameras are cheap and everywhere, so vision is the kind of AI you meet most often in public places. Recognising it helps you understand what is being collected about you and why.

Key terms

Application
A practical use of a technology to solve a real problem.
Optical character recognition
Turning a picture of text into text a computer can edit and search.
Number plate recognition
Reading vehicle number plates automatically from camera images.
Medical imaging
Pictures taken to look inside the body, such as X-rays and scans.
Surveillance
Watching people or places closely, often with cameras.
Assistive technology
Tools that help people with disabilities do everyday things.

Explained in detail

Reading: the quiet workhorse

The most widely used vision task is one of the least noticed: reading text from images. Post offices use it to sort letters by pin code. Banks use it to read cheques and identity documents. Phone apps use it to scan a printed page into text you can edit, or to read a visiting card into your contacts. Libraries and archives use it to turn old books and land records into text that can be searched. Without this, the paper records of the last few centuries would stay locked on their shelves.

Sorting and grading

Vision is used wherever things must be sorted by appearance. Machines in rice and pulse mills photograph every grain as it falls past a camera and use puffs of air to remove discoloured grains and small stones. Fruit packing lines grade mangoes and apples by size, colour and blemishes. Recycling plants separate plastic, paper and metal. These systems handle tonnes of material an hour with a consistency no person could maintain.

Health beyond the eye

Besides screening eyes for diabetic damage, vision models examine chest X-rays for tuberculosis, mammograms for early signs of breast cancer, and microscope slides for cancer cells and malaria parasites. In each case the model's job is to make sure that nothing is overlooked and that urgent cases are seen first. Studies of breast screening have found that a doctor working with an AI second reader can match the accuracy of two doctors reading independently, which matters wherever specialists are scarce.

Roads, crowds and public places

On roads, cameras read number plates to issue fines for jumping signals, measure the flow of traffic and detect riders without helmets. At very large gatherings, cameras linked to vision models estimate how dense the crowd is in each area, so that officials can redirect people before a crush develops. These uses can save lives. They also mean that movements in public are recorded on a scale that was impossible before, which is why they need clear limits on what is stored and who may see it.

Sight for those without it

For people who are blind or have low vision, a phone with a vision model is a remarkable tool. It can read a medicine label, say which currency note is in the hand, describe a photo sent by a friend, tell whether the lights in a room are on, and recognise the faces of family members who have agreed to be saved. Tasks that once needed another person's help can be done alone.

Four questions for any camera system

QuestionWhy it matters
What is the benefit, and to whom?A system should solve a real problem, not exist because it can
Do the people in view know and agree?Consent and notice are basic fairness
How accurate is it for every group?Errors fall unevenly when training data was uneven
What is kept, for how long, and who can see it?Stored images can be misused long after they were taken

How it works

FieldWhat the camera seesWhat the model does
MedicineX-rays, eye photos, skin photosFlags signs of disease for a doctor to review
FarmingLeaves, fields from drones and satellitesNames a disease, estimates crop health
TrafficRoads and junctionsReads number plates, counts vehicles, spots helmets
ShopsShelves and checkout countersNotices empty shelves, recognises products
SportThe ball and the playersTracks the ball's path for umpiring decisions
AccessibilityWhatever a phone camera points atDescribes the scene aloud for a blind user
DocumentsPrinted and handwritten pagesTurns the picture of text into editable text

A closer look: saving eyesight

Diabetes can damage the back of the eye and cause blindness, but it can be treated if caught early. There are too few eye specialists to examine everyone at risk. Vision models trained on many thousands of labelled eye photos can flag the signs of damage, so that a health worker can take a photo at a local clinic and the patients who need a specialist are found sooner. Such systems have been studied and used in Indian eye hospitals.

Try it yourself

A camera helps an umpire

A bowler appeals for leg before wicket. The batting side asks for a review.

Vision hunt

Over one day, look for computer vision in the places you go.

  1. List every camera you notice: at home, on the road, in shops, at a station.
  2. For each one, guess: is it only recording, or is a model reading the pictures?
  3. Write what you think it is looking for, such as number plates or the count of people.
  4. Mark each use as helpful, worrying, or both.

Think about it

  • Which use surprised you most?
  • Was there a sign telling people about any of the cameras?

Common mistakes

  • Thinking every camera uses AI. Many only record video for a person to watch later.
  • Thinking the model makes the final decision in medicine or sport. A qualified person reviews it.
  • Ignoring privacy because the purpose sounds helpful.

Check yourself

Question 1

1. What do vision systems in medicine usually do?

Question 2

2. Turning a photo of a printed page into editable text is a job for…

Question 3

3. Why is computer vision so common in public places?

Summary

  • Vision helps wherever careful, fast, repeated looking is needed.
  • Medicine, farming, traffic, shops, sport and accessibility all use it.
  • The model looks; a person decides; privacy still matters.

Finished reading?

Sources: AI4K12: Five Big Ideas in AI · Elements of AI (University of Helsinki)