We introduce an efficient, image formation model-based algorithm that extends super-resolution fluorescence localization to include orientation estimation, and report experimental accuracies of 5 nanometers for position estimation and 2 degrees for dipole orientation estimation.