This paper describes a system which locates and grasps doughnut shaped parts from a pile.The system uses photometric stereo and binocular stereo as vision input tools.