Computer vision is the branch of AI concerned with giving machines the ability to interpret visual information from the world, whether that is a photograph, a video stream, or live camera footage. The goal is to go beyond simply storing pixels and instead extract meaning, such as identifying objects, reading text, or tracking movement.
How a Computer 'Sees' an Image
To a computer, an image is just a grid of numbers representing colour and brightness at each pixel. A computer vision model, typically a type of neural network called a convolutional neural network, scans across this grid looking for small patterns like edges and textures in its early layers, then combines those simple patterns into more complex shapes in later layers, eventually recognising complete objects such as a face or a car.
Common Applications
- Facial recognition used for phone unlocking and photo tagging.
- Quality control cameras on manufacturing lines that spot defective products.
- Medical imaging tools that help flag potential abnormalities in scans.
- Optical character recognition that converts scanned documents into editable text.
- Object detection in self-driving cars to identify pedestrians, signs, and other vehicles.
Computer vision has matured rapidly over the past decade, moving from research labs into everyday consumer products. Even so, it still struggles with unusual lighting, occlusion, and situations that differ significantly from its training data, which is why human oversight remains important in high-stakes applications.