Bounding Box

A bounding box is essentially a rectangular outline drawn around an object or a region of interest within an image. This technique is common to annotate images for machine learning projects. It is mainly employed in the field of computer vision for tasks like object detection and image classification.

Bounding boxes is one of the simplest and fastest ways to localize an object in a data labeling project.

To create a bounding box, the annotator or labeler draws a rectangle around the object or region of interest in the image. It is using data labeling tools that we create this rectangular boundary. Then, we typically define it by two sets of x and y coordinates.  

Bounding boxes are used to annotate a variety of objects or regions in images, such as people, animals, buildings, vehicles, and many others. Often combined with other types of annotations (or labels), they provide additional information about the objects or regions represented. Two examples of different annotations are classes (to identify an object, such as “apple”, “pear”, “orange”) and attributes (to add object-specific details, such as maturity level, occlusion, etc.).

Finally, sometimes it is interesting to rotate bounding boxes to better localize some objects. It is a feature in some labeling tools named “oriented bounding box”.

Synonyms: Bbox; BB