In this work, standard machine learning techniques such as simple discriminative type classifiers (Logistic Regression) and generative type classifiers (Gaussian Mixture Model) are implemented [1], regarding the segregation of samples based on various stages of corrosion. The performance of the classifiers is evaluated using a histogram of misclassifications. Surface corrosion has a degrading effect on the material and presents itself normally as exfoliation or surface pitting or a combination of both. Corroded materials, e.g., in pipelines and reformer tubes can pose big threats to the environment, if corrective measures are not taken. Thus it is very important and necessary to detect and characterize the corroded regions. In the characterization of different states of corrosion several NDT techniques can be used [2]. In the case of electrically conductive materials such as reformer tubes, where austenitic or weakly magnetic steels are widely used, eddy current testing is the preferred method and it was used to characterize the state of corrosion of stainless steel plates which were divided in two groups: stage 1- when the corrosion is very small and causes no big threat to the structural integrity of the sample and stage 2 – for samples almost degraded with heavy corrosion. The induced eddy currents are perturbed due to the change in electrical conductivity produced by the corroded regions. This in turn affects the electromagnetic coupling of the coil to the sample and thereby affecting the coil impedance. This change in impedance is observed using an impedance analyser for various positions of the coil over the test specimen. Figures 1 (a) and (b) depict the change in the magnitude and angle of the coil´s complex impedance, respectively, for a line scan performed over the test sample.