Assessing the similarity of cortical object and scene representations through cross-validated voxel encoding models