Deep Learning is the key for the recent breakthrough in predicting protein structures from sequence information, with accuracy close to experimental resolution. This success will likely revolutionise drug development and biotechnology, as a consequence of the bold statement that ”structure determines function”. In contrast, the prediction of 3d RNA structures by Deep Neural Networks lags behind, mainly because of the low number of RNA only structures that are available for machine learning through the Protein Data Bank (PDB). In this ongoing project we use advanced data augmentation strategies to learn the mapping from RNA sequence to 3d structure.