The submission has received two positive and two negative reviews. The post-rebuttal discussion has not lead to convergence, and the opinion of the reviewers remain split. The concerns of the "negative" reviewers are: 1) The application is too niche (R1). However, the topic of the paper falls into NeurIPS call for papers, as it is related to low-level computer vision, compressed sensing, deep neural architectures. 2) The comparison to  may be invalid since the qualitative performance of  in the submission seems to be considerably worse than in the original paper. The authors rebut that the results in  were cherry-picked and that they use the code from , while fixing the parameters. There is a quantitative comparison with  suggesting that the new method is better. In the absence of evidence to the contrary, the rebuttal seems to be plausible. 3) There are other prior works, to which comparisons should have been implemented. The authors rebut that those methods take multiple images and are hence not comparable to the proposed one. 4) The comparison in Table 1 is misleading. This is indeed a concern, but does not constitute grounds for rejection in the opinion of the area chair. 5) The runtime comparison is misleading since CPU and GPU runtimes are compared. Again, this is a valid criticism, but it can be clarified in the final version without major revision of the paper and its claims. Overall, the negative reviews do not seem to provide grounds for rejection. Given the support of the two positive reviewers, the suggestion of the area chair is to accept. In the final version, the authors should incorporate the feedback of the reviewers. In particular, the comparison in Table 1 and Figure 4 should spell very clearly that the compared methods are not truly comparable and that the comparison is performed due to the lack of more similar prior art. Table 1 caption must be expanded. Furthermore, as promised in the rebuttal, the pros and cons of the HRD reconstruction approaches should be discussed in the related work section. The claim about 120x speedup should be revised and spelled more accurately.