Untrained Metamaterial-Based Coded Aperture Imaging Optimization Model Based on Modified U-Net
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn this paper, an untrained deep radar-echo-prior-based MCAI (DMCAI) optimization model is proposed to address the difficulty in collecting paired training data. The simulation and experimental results demonstrate the effectiveness of the proposed method. However, there are some problems in my opinion:
1) This paper proposes an untrained DMCAI optimization model. I want to know how to understand “untrained”? In fact, in the experiments, the proposed method still uses some training dataset.
2) This paper should provide a detailed description of the process of generating training data.
3) In this paper, the authors make some modifications to the classical U-Net architecture. But I do not see some significant changes compared to the classical U-Net architecture. In addition, you do not use the skip connection in the advanced U-Net structure, which should be validated through ablation experiments.
4) In the experimental section, the method proposed in [18] should be used as a comparative method.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsMy comments are in the file
Comments for author File: Comments.pdf
English is fairly good, some typo errors.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsDear authors,
After the review process of your manuscript, I would like to share some impressions that I found relevant.
- First, it presents a novel method that combines deep learning approaches to reconstruct target images, supporting its effectiveness with experimental results, that confirm its value.
- Second, one important aspect of your method is it overpassed other methods (such as SBL and TVAL3 algorithms). I am wondering if their implementation is free available (code), or do you make an own implementation of them? I think this is very relevant.
- Third, I consider the discussion section should be extended, maybe discussing on parameters values impact over the final images; also why the SSIM metric was not included in figures for comparisons, etc.
- Four, the M value is a parameter? because Fig. 5 mentioned it, but it was not clearly defined previously. The Line 273 indicates that M meaning measurements, but much after of being mentioned. Also, why do you choose the M=0.5N value, to evaluate the effect of noise on the performance of proposed method?
It is all by now, I hope this help you to improve a bit your interesting manuscript.
Best regards,
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for Authorspublish as it is