Object detection goes beyond labelling a whole picture; it finds where specific things are within it and draws a box around each. The system learns from many labelled images what each object tends to look like across angles, lighting and partial views.
It powers everything from sorting photos to guiding machines that must act on what they see. Its weak spots are unusual conditions and objects unlike anything in its training, where it may miss things or confidently mislabel them.