Computer vision means the part of AI that lets computers make sense of pictures and video: spotting objects, faces, text or movement. It is what lets your phone find all your photos of dogs, or lets a car notice a pedestrian.
A camera only records light, the way a tape recorder only records sound. Computer vision is the step after: like a person who looks at the photo and says “that is a bus, and that is a stop sign”. The camera sees; computer vision tries to understand what it sees.
In a sentence: “The parking app uses computer vision to read my number plate as I drive in, so there’s no ticket to lose.”
What the word means
The plain point: it is AI for anything visual.
IBM defines it in those terms: “Computer vision is a subfield of artificial intelligence (AI) that equips machines with the ability to process, analyze and interpret visual inputs such as images and videos.”1
Its common jobs have plain names:
- Classification: what is this picture of?
- Object detection: where are the things in it?
- Segmentation: which exact pixels belong to each thing? This is how a tool cuts out a person from a photo.
- Text recognition (OCR): reading printed or written words in an image.
- Facial recognition: matching a face to a person.
A simple example
The plain point: it learns to see from many labelled pictures.
IBM gives a hospital example: to build a system that spots pneumonia, staff first gather chest X-rays and mark each one.3 Those marks are labels, and learning from them is supervised learning.
IBM adds that the pictures must be numerous and varied.3 Where real images are hard to get, builders sometimes make synthetic data.
Where you will meet it
The plain point: it is already in tools you use.
- On your phone and computer: removing a photo’s background is segmentation at work. How to remove a background with AI walks through it.
- In chatbots: when you upload a photo and ask about it, the model is doing a kind of computer vision. See multimodal meaning.
- In video tools: models that find and follow objects in a clip. SAM 3.1 is one example, from Meta.
- In industry: IBM lists self-driving vehicles, medical imaging, factory quality checks, shops and robots.3
Why it matters to you
The plain point: it can be wrong, and more wrong for some people.
A 2018 study called Gender Shades found that commercial systems that guess gender from a face photo made far more mistakes on darker-skinned women than on lighter-skinned men: error rates of up to 34.7%, against at most 0.8%.4 AI bias meaning goes into more detail. If a system that looks at faces makes a decision about you, it is fair to ask how well it was tested on people like you.
Computer vision is a pattern-matcher for pictures. Check its work on the hard cases: hair, fingers, low light, unusual angles, and faces it saw few of in training.
Often confused with
- Image generation: making new pictures from text. IBM counts it within the field, but it is the reverse job: creating images rather than reading them. See diffusion model meaning.
- Deepfakes: fake video, photos or voices of real people, made to deceive. That is a use of image-making AI, not what the term computer vision means. See deepfake meaning.
More words like these: the AI glossary and AI explained for complete beginners. How image recognition stopped counting as “intelligence” once machines could do it is told in a short history of AI.
Questions people ask
Is facial recognition computer vision?
Yes. Matching faces is one of the field’s standard jobs, alongside reading text in images and spotting objects.2
How does computer vision learn to see?
From examples. People collect lots of pictures, label what is in each one, and the system learns the patterns. IBM stresses that the pictures need to be many and varied.3 See training data meaning.
Related on this site: AI for People With Disabilities · What Is a Vision-Language Model?
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is computer vision?
- A camera that records light very sharply
- The part of AI that lets computers make sense of pictures and video
- AI that makes brand-new pictures from text
Show the answer
Answer: The part of AI that lets computers make sense of pictures and video. Computer vision is AI for anything visual: spotting objects, faces, text or movement in pictures and video.
2. How does a computer vision system learn to spot something, like pneumonia in X-rays?
- A doctor explains the rules to it in one sentence
- It copies answers from other hospitals
- From many varied pictures that people have labelled
Show the answer
Answer: From many varied pictures that people have labelled. People gather lots of varied pictures, label what is in each one, and the system learns the patterns from those labels.
3. What does the page suggest checking computer vision on?
- The hard cases, like hair, fingers, low light and faces it saw few of
- Only clear, well-lit pictures
- Nothing, because it does not make mistakes
Show the answer
Answer: The hard cases, like hair, fingers, low light and faces it saw few of. The takeaway calls it a pattern-matcher for pictures and says to check its work on the hard cases.
Related terms: Attention · Checkpoint · Clustering · Constitutional AI · Distillation · Emergent abilities
- 1IBM Think, What Is Computer Vision?, Rina Diane Caballar and Cole Stryker, 17 Sep 2025, read at source 1 Oct 2026“Computer vision is a subfield of artificial intelligence (AI) that equips machines with the ability to process, analyze and interpret visual inputs such as images and videos.” ↑
- 2IBM Think, What Is Computer Vision?, read at source 1 Oct 2026, lists tasks including “Image classification, object detection, image segmentation, object tracking, scene understanding, facial recognition, pose estimation, optical character recognition, image generation, visual inspection.”↑a↑b
- 3IBM Think, What Is Computer Vision?, read at source 1 Oct 2026“Hospitals will need to compile a dataset of chest X-ray scans and correctly label or annotate each scan as either normal or signifying pneumonia.” ↑a“datasets must also be sufficiently large and diverse enough for computer vision algorithms to produce accurate results.” ↑bapplications include “autonomous vehicles”, “healthcare”, “manufacturing”, “retail and e-commerce” and “robotics”. ↑c“datasets must also be sufficiently large and diverse enough for computer vision algorithms to produce accurate results.” ↑d
- 4Buolamwini & Gebru, Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification, Proceedings of Machine Learning Research 81, 2018, read at source 13 Oct 2026“darker-skinned females are the most misclassified group (with error rates of up to 34.7%). The maximum error rate for lighter-skinned males is 0.8%.” ↑
The comparison on this page (a camera as a tape recorder, a person naming what is in the photo) is this site’s own way of explaining, not a claim from this source.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK