Learning grasping affordances from local visual descriptors | IEEE Conference Publication | IEEE Xplore