Deals with one of the oldest problems in Computer Vision: to recover the 3-D geometric and kinematic structures of the world from two images, and to recognise a 3-D object in a cluttered scene from one or several views of this object in a different setting.