The white shape shows the original linear transformation, which is a composite of scaling, rotation and translation (3D skewing). The other two shapes show what happens when the factors of an eigendecomposition are applied as linear transformations. The eigendecomposition is Q*D*Q^(-1). The orange shows the effect of Q*D, while the green of Q^(-1).
This is the GitHub repo of the project: https://github.com/bolner/Visual-Matr...