PASCAL-Part is a set of additional annotations for the PASCAL VOC 2010 dataset that provides segmentation masks for each body part of the detected object, and contour annotations for objects with non-uniform parts like boats.

In addition, PASCAL-Part can also be used for semantic part segmentation of the human body, including portraits in all images (1,716 for training and 1,817 for testing), providing pixel-level accurate annotations for six parts of the human body (i.e., head, torso, forearm, hind arm, thigh, and calf).