NeurIPS 2020

Self-Adaptively Learning to Demoiré from Focused and Defocused Image Pairs


Meta Review

Two of the reviewers recommended accepting this paper and a third recommended rejection. The two main concerns of the negative reviewer were the experimental comparisons and the importance of Moire patterns. The authors convincingly addressed the first point in their rebuttal, and in the discussion the reviewer acknowledged this fact, yet stood by the assertion that Moire patterns are not of major importance in modern cameras. I tend to agree with the other two reviewers: while Moire patterns are somewhat of a niche phenomena, they are still interesting and the approach here is novel and interesting. So I am happy to recommend accepting the paper but urge the authors to do a better job of motivating the problem to the general NeurIPS audience. Some of the material in the rebuttal can be useful to add to the revised paper in this respect.