SERNet: Squeeze and Excitation Residual Network for Semantic Segmentation of High-Resolution Remote Sensing Images
and Mahmoud Emam 4
Round 1
Reviewer 1 Report
(1) The author proposed SERNet for Semantic Segmentation of High-Resolution Remote Sensing Images, while the related works are mainly based on the study of computer vision, such as the network and attention mechanism. Many studies on land-cover segmentation were proposed in recent years, similarly focused on the problems of the multi-scale objects and accurate boundaries, etc. Maybe more research on remote sensing imagery should be investigated in the part of related works.
(2) The author conducted experiments on the ISPRS Vaihingen and Potsdam datasets, but most of the compared methods are classical segmentation network which is not designed for remote sensing imagery. It is advised to compare more related works related to remote sensing which conducted experiments on these datasets for a fair comparison.
(3) Compared to Table 3, Table 4 seems to miss a result of SERNet (a-DSM) on the ISPRS Potsdam dataset?
(4) From Figures 6 to 9, the results are not aligned. The marked box seems a little casual. It is advised to make a uniform style for the result visualization.
(5) The title focused on remotely sensed imagery segmentation, while the content is more focused on the segmentation of vegetation categories. May the title be slightly adjusted to be consistent with the article.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Dear authors,
Upon carefully reviewing and evaluating your manuscript, I’m recommending it for revision. This is an interesting topic, but I detect structural deficiencies in its sections. In this regard, I have detailed my suggestions, which I believe will improve the overall quality of the manuscript content. I hope these comments are useful to you.
1. Please move Figure 1 to be displayed at the top of page 3 instead of the middle of page 2. I believe that it is an issue with the Latex formatting, but preferably figures should be displayed after they are mentioned in the text, and not before it. (Also, why the lines are not numbered? This makes it difficult to provide revision information). Also, Figure 2 is before its introduction on the next page. Please consider altering the appearance of each figure to be after the paragraph that they are mentioned.
2. The 3rd paragraph of the introduction feels disconnected. You were talking about the necessities of implementing learning mechanics into CNNs, and suddenly you wrote “In addition, the increase of carbon…”. I fail to see how this paragraph can relate to the 2nd, therefore this “In addition…” feels awkward. The whole part “In addition, the increase of carbon content in the atmosphere leads to the rise of global temperature and affects the regional and global climate conditions [13]. Vegetation absorbs carbon through the normal photosynthetic process, provides storage for carbon in the ecosystem, and thus effectively controls the carbon content in the atmosphere [14].” Appears to derivate from another segment. Please consider removing it or moving it to a more appropriate place.
3. In general, the introduction and related works sections are poorly handled. I believe that the authors should merge these two sections and present their writing in a more sequential manner. In this format, there’s too little information on each subsection to contextualize as a literature revision of each topic, so it is better to remove the subsections, merge the two main sections and present these paragraphs in a more logical order.
4. The method section, where the network is described, is fine. But the “results” subsections, 4.1 and 4.2 are protocol setup and data organization. These sections should be in the method main section, not results since they are describing the data preparation and the details regarding network implementation, training, and validation procedures.
5. I’m having trouble identifying the “Tree” class in Figure 9 original imagery. Is it possible to implement a different composition or a contrast variation to help readers discern it?
6. The discussion section is fine, but this whole paper reads more like a Technical Note instead of an actual Article. I advise the authors to improve their discussion section.
7. In the general sense of this manuscript, the English language needs to be improved; there are grammar and sentence structure errors in the text. I advise a careful examination in a subsequent read.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
There is no comment for the revised manuscript.
Reviewer 2 Report
The authors have considerably improved this version. I believe it is appropriate for publication now.
