In this paper, we attempt to evaluate the shape of a surface flaw including its horizontal position and the located surface by biaxial Magnetic Flux Leakage Testing (MFLT) using an amorphous MI sensor and a Neural Network (NN). The specimen is a magnetic material subjected to the magnetic field, and the Magnetic Flux Leakage (MFL) occurs near the flaw. We measure the biaxial MFL, i.e., the tangential and the normal components of MFL by an amorphous MI sensor. The amorphous MI sensor has the wide measurement range, high sensitivity and high spacial resolution, so that it is suitable for precise quantitative evaluation of the flaw by MFLT. Initially, we pre-process the measured biaxial MFL by Regression Analysis Method (RAM) to extract MFL parameters. Subsequently, NN is used to infer the dimension of the cross section of the flaw including its horizontal position and the located surface from the MFL parameters. By repeating a similar process along several measurement lines parallel to the specimen surface, we can identify the three-dimensional shape of the flaw. In this paper, we first evaluate three-dimensional shape of a parallelepiped flaw in SS400 specimen as the simplest case. Secondly, we consider extending the evaluation method to an oblique flaw based on the two-dimensional magnetostatic analysis by use of Finite Element Method (FEM). The results show that the three-dimensional shape of a parallelepiped flaw and the two-dimensional shape of an oblique flaw can be evaluated with good accuracy. © 2009 Springer-Verlag Berlin Heidelberg.