U-Net for Taiwan Shoreline Detection from SAR Images
Round 1
Reviewer 1 Report
I have two main concerns.
First, I do not feel as comfortable with the topic as I thought I would. Second, and perhaps more importantly, I find the document hard to read. In many parts of the document I have the impression that not enough effort has been made by the authors to make the reading pleasant and, especially, easy to follow (e.g., Figure 5 should clearly be the first to appear; note also a "Table 2" appearing above the figure). On the other hand, the writing is poor and this adds to the difficulty of reading: it is enough to read the introduction to realize this. In short, I believe that the work is not fit to be reviewed to discuss its scientific value (about which, as I said, I may not be the best person to give an opinion).
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper shows a lot of work, good results and the solution is very relevant. The work could be improved by incorporating more novel ideals. Nevertheless, the work is sufficient for publishing.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
n this paper, UNet model is used to detect the coastline, and the convolution layer of UNet model is added with BN module. At the same time, several experiments are carried out to verify the practicability of this method. However, the method in this paper mainly uses UNet model for a small amount of optimization, which is lack of innovation, and systematically improvement of UNet according to the characteristics of the coastline is not found. In addition, whether the method in this paper is superior to Res-UNet and Deep-unet has not been proved by comparative experiments.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
The paper presents a trending subject and it is well structured but requires some work in order to be ready for publishing. English syntax and grammar needs to be improved throughout the paper. Some examples are given below:
1. Line 36 - "greenhouse gas is lead" should be "has lead".
2. Line 39 - "reason for the" instead of "reason of".
3. Lines 60 and 61 - Please rephrase to improve syntax.
4. Please don't use "the" before study [x].
5. Line 96 "in order to classify" instead of "classified"
6. Please specify figures in the correct order in the paper. You have started with Figure 5 instead of Figure 1. Include the Figure in the paper as soon as possible after the reference in order to improve readability.
7. Please don't use colloquial language in the paper, such as "by the way", present on line 250
8. You have an isolated Table 2 heading on line 342.
Some other issues need to be addressed regarding the scientific content:
1. Line 206. Why haven't you used the Copernicus portal to obtain the Sentinel-1 data. Also, why haven't you extended the study until the present and stopped in 2019?
2. Can you please provide the results of the training without the BN layer to better highlight your contribution and how it improves the results.
3. Due to the small analysis window and low resolution of the data, no major, natural, changes could have been detected (in the absence of any natural disaster). Why haven't you just concentrated on human made coastline changes.
4. Please provide some comparison regarding the performance of your method compared to other Sentinel-1 data coastline detection methods.
5. The overall results are quite lackluster. Why didn't you try to improve your results through methods such as data fusion, maybe with some ground truth data.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
There is not much innovation in this method, only the application of existing methods.
Reviewer 4 Report
The authors have addressed all of the raised issues. The quality of the presentation has greatly improved. Some small typos are still present but they can be identified by the authors/editors.
