Right now the SVD of the covariance matrix will yield the optimal rotation to align the shapes. Need to research whether this rotation is orthogonal or whether it can lead to flipping of the models. (Seems the scale is positive definite, so how else could we flip?!) Once this is established, add a boolean **kwarg to the Procrustes constructor so we can choose to allow flipping or not.
Right now the SVD of the covariance matrix will yield the optimal rotation to align the shapes. Need to research whether this rotation is orthogonal or whether it can lead to flipping of the models. (Seems the scale is positive definite, so how else could we flip?!) Once this is established, add a boolean **kwarg to the Procrustes constructor so we can choose to allow flipping or not.