Annotations are markings added to an image. In object detection, they are bounding boxes; in segmentation tasks, they can be polygons or segmentation maps. These markings serve to teach the model the ground truth, guiding it to recognize and classify objects accurately.
