SBD Dataset is a semantic boundary dataset that contains annotations for 11,355 images from the PASCAL VOC 2011 dataset, all of which are based on Amazon Mechanical Turk. Conflicts between segmentations are manually resolved. In addition, each image consists of category-level and instance-level segmentations and boundaries for 20 object categories in the PASCAL VOC 2011 challenge.

This dataset was released by the University of California, Berkeley in 2011, and the related paper is "Semantic contours from inverse".