To a computer, an image starts as nothing more than a grid of numbers representing color at each pixel. Computer vision is the field of building models that turn that grid of numbers into something meaningful — a face, a stop sign, a tumor, a defect on a factory line.
Most modern computer vision systems rely on convolutional neural networks, a design that scans small patches of an image looking for simple patterns like edges and textures, then combines those patterns in deeper layers to recognize increasingly complex shapes and objects.
The applications are everywhere once you start looking: quality control cameras on assembly lines, medical imaging tools that flag areas for a radiologist to review, agricultural drones that spot crop disease early, and the object-detection systems that let a self-driving car tell a pedestrian from a parked car.