Turning pixels into meaning
To a computer an image is just a grid of numbers. Computer vision is the work of turning that grid into something meaningful: is there a face here, what breed is this dog, where is the tumour, is the lane clear. Tasks range from classification (what is this) to detection (what is where) to segmentation (which pixels belong to which thing).
Convolutional neural networks made modern vision possible, and more recently Transformers have moved in here too — the same architecture that powers language. It is a flagship application of deep learning, and the visual counterpart to natural language processing.
Now part of generative AI
Vision isn't only about recognising images — generative models create and edit them too, and multimodal systems handle text and images together.