Our main contribution is to use contrastive learning for creating image-like embeddings suitable for registration, and