Your phone unlocks when it sees your face. A factory line stops when it detects a defective part. A radiologist gets a second opinion from software that has analyzed millions of scans. All of this is computer vision: the field that enables computers to process and understand visual information the way humans do, or in some cases, better.
It starts with pixels
To a computer, an image is a grid of numbers. Each pixel has a numerical value representing its color and brightness. Computer vision is the process of turning those numbers into meaning, and identifying what's in an image, where things are, and what's happening.
This used to require hand-crafted rules: "if there's a round shape of this color range in this region, it's probably a face." Modern computer vision uses convolutional neural networks – a type of deep learning model that learns to detect features like edges, textures, and shapes automatically, by training on large labeled image datasets. The result is systems that can identify objects, scenes, and patterns far more reliably than rule-based approaches.
Object detection – finding things in images
Object detection identifies what objects are present in an image and where they are, drawing a bounding box around each one. You've seen this every time a camera automatically focuses on faces in a group photo, or when a security camera flags an unauthorized person entering a restricted area. In retail, the same technology tracks which products customers pick up and put back on shelves. In warehouses, it guides robotic arms to the right items.
Image recognition – classifying what's there
Where object detection finds and locates, image recognition classifies. Given an image, it answers: what is this? Dermatology apps that analyze skin lesions, manufacturing systems that sort products by visual quality, and content moderation tools that automatically flag inappropriate images all rely on image recognition. It's the reason you can use the search function in your gallery, successfully finding all images featuring a dog or an ice cream cone.
Semantic segmentation – understanding the full scene
Segmentation goes further than detection: instead of drawing a box around an object, it identifies every pixel that belongs to it. Semantic segmentation labels every pixel in an image with a category: road, car, pedestrian, building. This level of detail is what self-driving cars need to navigate safely – knowing not just that there's a person nearby, but exactly where their outline is and how much space they occupy.
Optical character recognition – reading text in images
Optical character recognition (OCR) converts text that appears in images into editable digital text. Every time you photograph a document and your phone lets you copy text from it, or a border crossing automatically reads a passport, OCR is at work. It's also what allows banks to process cheques automatically and what makes scanned archives searchable.
Where it's heading
Computer vision is one of the fastest-moving areas of AI. The same underlying technology that reads license plates and counts warehouse inventory is now being applied to medical imaging by detecting tumors in scans earlier than human specialists, analyzing retinal images for early signs of diabetes, and counting cells in pathology slides. The pattern is consistent: take a task that requires a human to look at something carefully, and ask whether a trained model can do it faster, more consistently, or at a greater scale.
For a full reference of AI terms used in this article and beyond, see the AI Dictionary →