The Final Project of the lecture Generative Neural Networks (GNN) in Heidelberg University. This is an idea to solve the competition on Kaggle: https://www.kaggle.com/competitions/gpr-max-deep-learning-challenge-1-gdlc-1. We combine Normalizing Flows with U-Net architectures for one-shot multi-offset Ground Penetrating Radar (GPR) data.