This paper introduces a framework for data augmentation in subspace clustering. The method is well explained and back by thorough and convincing experiments. The main drawback of this work is that it lacks motivations to justify the need for this method, and this may greatly diminish the impact of the work if not clarified in the next iteration.