[Paper Review] SimCLR: A Simple Framework for Contrastive Learning of Visual Representations