Manipulation-Independent Representations(MIR) for Successful Cross Embodiment Visual Imitation