There have been many efforts put in by the machine learning (ML) community to improve classification. In this study, we develop a new conjugate gradient Vector Field (VF) \(\overline{\nabla }\hat{\psi }\) . The main features of a VF are its singularities and trajectories. The gradient VFs generate three kinds of singularities, while the conjugate gradient VF \(\overline{\nabla }\hat{\psi }\) has seven kinds of singularities, each with a different shape. Embedding VF into an image incorporates its features into the image. This way, we extend the set of image features with VF features which improve the classification statistics. In the present study, we validate that embedding the new VF \(\overline{\nabla }\hat{\psi }\) in image databases increases the classification metrics of ML classifiers. For this validation, we used five different NNs, which are trained with three sets of original image databases (COIL, digit MNIST, and Fashion MNIST) and five sets of image databases with embedded VFs \(\overline{\nabla }\hat{\psi }\) , \(\overline{\nabla }\hat{\phi }\) , and \(\overline{\nabla }\hat{u}\) . At the end, we show results confirming that the image databases with embedded VF increase ML classification metrics.