Computer vision
What is Computer vision?
The field of AI that lets machines interpret images and video: recognising objects, reading text, spotting defects, tracking movement. It is the technology behind quality inspection on a production line, medical image analysis and document scanning. Where language models read and write, computer vision sees.
Why it matters
Computer vision delivers its clearest wins in physical operations, where a camera can watch a process no person has time to inspect fully. The catch is that these systems are unusually sensitive to the conditions they run in. A model trained on well-lit photos can fail on a dim factory floor, a new camera angle, or a product variant it never saw, and unlike a software bug the cause is often physical: a smudged lens, a shift change, a supplier’s redesigned packaging. That makes the deployment environment part of the system, not a detail. Lasting value depends less on the cleverness of the model and more on whether real-world conditions stay close to what it was trained on, and on catching it when they drift.
In practice
A manufacturer installs a vision system to spot defects on the line, and it performs well for months until reject rates suddenly spike. The model has not changed; a new overhead light has, altering how surfaces reflect. Once the team learns to treat lighting, camera position and product changes as things that break the model, they add regular checks against fresh samples. The technology worked; keeping the conditions stable was what kept it working.
