What computer vision actually sees

A computer vision system does not see a scene the way a person does. It sees a grid of numbers, and every stage of the pipeline exists to turn that grid into something a business can act on.

The first stage is detection: locating the regions of a frame that matter, whether that is a vehicle at a gate or a face turned toward a whiteboard. The second stage is classification and tracking, which gives those regions identity and continuity across frames.

The last stage is the one organizations actually pay for: turning tracked detections into a metric, an alert, or a dashboard that a operations or academic team can use without ever looking at a video feed.