Biometrics fingerprint matching has been done with a heavily hand-tuned and designed process of classical computer vision techniques for several decades. This approach has led to accurate solutions for solving crimes today and, as such, little effort has been devoted to using deep learning in this domain. Exemplary embodiments disclosed herein leverage synthetic data generators to train a neural fingerprint enhancer to improve matching accuracy on real fingerprint images.