YOLO detects all objects in a frame in one forward pass, which makes it fast enough for live video. Successive versions (v8 through v12 and beyond) trade a little architecture for better accuracy and easier export to edge hardware, making YOLO the default production detector for many teams.