Representation Learning For Vision