Scene Representations For View Synthesis With Deep Learning