To fully exploit the predictive accuracy of advanced anisotropic yield functions, a large number of classical mechanical tests is required for calibration purposes. The Finite Element Model Updating (FEMU) technique enables to simultaneously extract multiple anisotropic parameters when fed with heterogeneous strain fields obtained from a single information-rich experiment. This inverse approach has the potential to mitigate the experimental calibration effort by resorting to a single, yet more complex experiment augmented with Digital Image Correlation. In this paper, we inversely identify the sought anisotropic parameters of two selected yield functions for a low carbon steel sheet based on the previously designed information-rich tensile specimen. The experimentally acquired strain field data is used to inversely identify the Hill48 yield criterion and the Yld2000-2d yield function, respectively. The results are compared with conventional calibration methods for both anisotropic yield functions. The inverse identification is then thoroughly studied using virtual experiments enabling to disentangle the effect of the material model error and the strain reconstruction error (DIC), respectively. It is shown that the material model error dominates the inverse identification of the Hill48 yield criterion. The reduced material model error for the Yld2000-2d yield function enables obtain inversely identified anisotropic parameters that are closer to the reference parameters. The paper clearly shows the importance of the predictive accuracy of the selected anisotropic yield function when applying inverse identification. Keywords: Anisotropic yield criteria; Material parameters identification; Heterogeneous mechanical tests; Inverse identification; DIC.