Advancing Physically Informed Autoencoders for DTM Generation
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThank you for submitting your research on transforming Digital Surface Models (DSMs) into Digital Terrain Models (DTMs) by combining remote sensing and deep learning (DL). Overall, the methodology is reasonable, but the structure of the article needs further adjustment. Here are my suggestions for each section:
1. (line 125-131) I suggest consolidating points three and four in the contribution section as they both relate to the innovative aspects of your methodology.
2. (line 132) The 'Related Work' section requires restructuring. The current version appears to be a mere compilation of previous studies, lacking a clear hierarchy and organization. Furthermore, it is recommended to provide a more comprehensive analysis of the limitations and gaps in the current body of research to justify the motivation.
3. Can you add a colorbar to the image of the data? For instance, in Figures 5, 6, 7, 8, 11, 13.
4. (line 298) In the experiments conducted using the OpenGF Dataset, it appears that only areas S4, S7, and S8 were selected for analysis. Could the authors please elaborate on the reasoning behind the choice of these specific areas?
5. (line 434) The ablation studies in your paper are comprehensive. However, they do not include a comparison with open-source state-of-the-art (SOTA) methods. To convincingly demonstrate the effectiveness of your proposed method, it is crucial to benchmark it against established techniques.
6. (line 585-586) The conclusion of your paper briefly touches upon the limitations of your method. However, it would be beneficial to have a more extensive discussion on this aspect.
7. In your manuscript, you describe your method as "physically-informed". Could you please provide more context or evidence in the paper to support this claim?
Additionally, there are several issues with the formatting and expression in your paper：
8. I noticed that your paper contains several instances of repeated ideas and excessively verbose expressions, such as line 62-63, line241-243, lines 390-393, line 420-427.
9. It seems like your manuscript has some noticeable formatting errors that need to be addressed. Specifically, you mentioned issues such as repeated figure captions (Figure 14 and 15), missing images (Figure 14 and 18), and incorrect figure numbers (line 298).
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper subject is appropriate for Remote Sensing as the authors provide DTM generation based on U-net with subtractive skip connection.
The idea is acceptable for the publication but the manuscript should be revised to provide more detail information on the design and the experimental results.
I think the paper needs modifications to be accepted.
Q1. On page 1, line 12, Quantitative summary is required about how much performance was increased over existing method. (e.g. RMSE, error rate, etc)
Q2. On page2, lines 68-80, quantitative information should be added on how efficient the proposed method is in terms of computation and memory usage compared to other methods. because you introduction mentioned about importance of optimization (”point-based methods require significant memory and processing cost ~”)
Q3. On page6, Figure 2, Figure 2 need more explanation, especially about Encoder and Decoder (e.g. convolutional layer size , pooling layer size, stride size, padding size, why the number of filter is chosen, etc.)
Q4. On page6, line 229, Explanation about activation function is also required
Q5. On page19, line 398, Model’s performance metric(precision, recall, F1 score) is required. Not only just about metrics in Quality Assessment Criteria
Q6. On page25, line 573, Author’s insight about why subtractive skip connections provide better results is required (e.g. visualization and comparison of feature maps under subtractive skip connections and concatenative skip connections)
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors1、It is suggested that a more specific description of the DTM generation model in this paper.
2、It is recommended to add the comparison experiment with other methods which mentioned in related work.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors have carefully addressed all my concerns to a satisfactory level, I have no further comment.
