Expected and predicted colour-magnitude diagram for the different training set we explored. Upper panels show the expected (BP-RP)₀ – M_G CMDs for a training sample composed only of stars that respect the parallax criteria presented in Sect. 2.2 (on the left), with the inclusion of K-giants (in the middle), and with the inclusion of both K-giants and BHBs (on the right). We only show the stars in common between the different samples, because the number of stars used to train the ANN are different from one training sample to another because of the different constraints used to create them. The lower panels show the corresponding CMDs with the absolute magnitude M_G predicted by the ANN. We can clearly see the improvement made by the inclusion of the different catalogues on the global shape of the CMD, especially from the K-giants.