What is Computer Vision?
Last updated · 1 min read
Pixels in, meaning out. A model is trained on millions of labeled images until it can flag "tumor-like shape", "person walking" or "scratch on this part" in images it has never seen — the same learn-from-examples idea as an LLM, applied to sight instead of text.
Where it works best: narrow, high-volume checks — radiology screening, factory quality control, license-plate reading. Where it struggles: context, intent and anything adversarial (a sticker can make a stop sign "invisible" to a model). In healthcare it increasingly works alongside doctors as a second pair of eyes — see AI in Healthcare.