Bounding box

A bounding box is a rectangular annotation drawn around a target object in an image or video frame. It’s one of the most widely used annotation types in computer vision, especially for object detection.

The coordinates of bounding boxes are typically defined by the top-left and bottom-right corners, or by the box center plus width and height. Each box usually includes a class label and may include extra fields such as occlusion level or truncation status.

Bounding boxes balance annotation speed and spatial accuracy, but they do not capture an object’s exact shape or contour.

Bounding box is the primary label format in benchmarks such as PASCAL VOC and COCO.