An Unsupervised Saliency-Guided Deep Convolutional Neural Network for Accurate Burn Mapping from Sentinel-1 SAR Data
Round 1
Reviewer 1 Report
Burn mapping is a trendy area, and many interesting studies exist. For example:
Giandomenico de Luca , João M.N. Silva and Giuseppe Modica. A workflow based on Sentinel-1 SAR data and open-source algorithms for unsupervised burned area detection in Mediterranean ecosystems,
Saygin Abdikan et all, Burned Area Detection Using Multi-Sensor SAR, Optical, and Thermal Data in Mediterranean Pine Forest,
end other.
The authors should highlight the novelty and benefits of their approach compearing with other modern results.
The primary significance of the study is questionable. The DCNN architecture is simple, and its development and evaluation are too weak points for publication in Remote Sensing.
The authors should think much better about the introduction and motivation part of their research.
The rest of the paper is organized fine unless the accuracy metrics should be described in detail, and there is no discussion section at all.
Author Response
Thank you for providing the opportunity to submit revised draft of manuscript of "An unsupervised saliency-guided deep convolutional neural network for accurate burn mapping from Sentinel-1 SAR data" to “Remote Sensing” journal. We appreciate the time and effort the editors and the reviewers have dedicated for suggesting great and valuable feedback.
We have put our attempt to completely correct the mistakes and improve quality of the manuscript.
Regards
Author Response File: 
Author Response.docx
Reviewer 2 Report
The manuscript presents an unsupervised saliency-guided deep convolutional neural network model to detect burned areas in Sentinel-1 SAR images. In addition to the high precision of the proposed method, it can be implemented unsupervised, enabling it to be used in diverse conditions when ground truth or a reliable training dataset is unavailable. The novelty is new and applicable, very detailed experiments are provided. The reviewer suggests its acceptance with minor edits. I have only some minor remarks:
1) The use of English in this paper needs to be moderately improved. All grammar and typing errors need to be corrected, such as L2 in line 206 and equation (8).
2) In equation (4) and (5), what are the definition of all variables, such as C and k?
3) Please give the detailed structure of the DCNN used in this article and the flow diagram of the proposed approach based on the SG-FCM-DCNN model.
4) For figure 2 (c)(d), what is the threshold?
5) How is the number of categories determined in Figure 3? And in line 232, “More than 8 clusters do not significantly improve results.”. What does the sentence mean?
6) Individual experimental data are somewhat lacking in validity. Could the author add more experiments?
Author Response
Thank you for providing the opportunity to submit a revised draft of the manuscript of "An unsupervised saliency-guided deep convolutional neural network for accurate burn mapping from Sentinel-1 SAR data" to the “Remote Sensing” journal. We appreciate the time and effort the editors and the reviewers have dedicated to suggesting great and valuable feedback.
We have put an attempt to completely correct the mistakes and improve the quality of the manuscript.
Regards
Author Response File: Author Response.docx
Reviewer 3 Report
The authors have proposed a framework based on SG- 19 FCM-DCNN. The manuscript is complete, and the authors try to prove the progressiveness of the algorithm through experiments. However, in general, the manuscript lacks innovation, and it is recommended to reject it. The comments are as follows
1. In the abstract, the authors said that the proposed algorithm needed fewer time to train the network, but the paper did not explain it, nor did it use experiments to prove it.
2. The paper does not use experiments to prove the progressiveness of the proposed algorithm. In the experimental part, computational complexity analysis experiment should be added.
3. The algorithm description is unclear and the logic is chaotic.
4. What is the specific structure of the algorithm? This is not specifically reflected in the paper.
Author Response
Thank you for providing the opportunity to submit a revised draft of the manuscript of "An unsupervised saliency-guided deep convolutional neural network for accurate burn mapping from Sentinel-1 SAR data" to the “Remote Sensing” journal. We appreciate the time and effort the editors and the reviewers have dedicated to suggesting great and valuable feedback.
We have put an attempt to completely correct the mistakes and improve the quality of the manuscript.
Regards
Author Response File: Author Response.docx
Reviewer 4 Report
The log-ratio of differences between areas before and after a fire (NBR) is what makes this study unique. Further there are standard steps of training deep neural network to identify clustered (using fuzzy c-means) salient regions based on NBR. The whole paper is a summary of the steps of the method and a comparison to other methods (which are only mentioned in the results) to show how much better it is to use remote sensing to find out how big burned areas are.
Some inconsitences of figures are: in figure 3 there are green colors of image representing burn areas, however clustering use also red themes.
The methodology of estimation and significance of precision (P), recall (R), and F1-score (F1) indexes are not explained in the paper, and only the overall accuracy (OA) of the chosen method proves the efficiency of the realized study.
Author Response
Thank you for providing the opportunity to submit a revised draft of the manuscript of "An unsupervised saliency-guided deep convolutional neural network for accurate burn mapping from Sentinel-1 SAR data" to the “Remote Sensing” journal. We appreciate the time and effort the editors and the reviewers have dedicated to suggesting great and valuable feedback.
We have put an attempt to completely correct the mistakes and improve the quality of the manuscript.
Regards
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors made a good job improving their manuscript. I have no further comments.
Author Response
Thanks for your valuable comments. We revised our paper based on your comments and attach the response letter.
Author Response File: Author Response.docx
Reviewer 3 Report
-The paper should be interesting ;;;
-it is a good idea to add a block diagram of the proposed research (step by step);;;
-it is a good idea to add more photos of measurements, sensors + arrows/labels what is what (if any);;;
-What is the result of the analysis?;;
-figures should have high quality. ;;;;;
-text should be formatted;;;;
-references should be formatted;;;
-what will society have from the paper?;;
-labels of figures should be bigger;;;;
-Is there a possibility to use the proposed research, methods for other topics, image processing, neural networks;;;
-please compare the advantages/disadvantages of other approaches, other neural networks;;;
-references should be from the web of science 2020-2023 (50% of all references, 30 references at least);;;DOI: 10.1016/j.eswa.2023.119508，10.1109/TGRS.2022.3202865，10.1109/TGRS.2022.3225902，10.1109/TGRS.2022.3201056
-please add some sentences about future work;;;
-Conclusion: point out what have you done;;;;
Author Response
Thanks for your valuable comments. We revised our paper based on your comments and attach the response letter.
Author Response File: Author Response.docx
Reviewer 4 Report
I have no further comments. I accept the text of the paper as corrected.
Author Response
Thanks for your valuable comments. We revised our paper based on your comments and attach the response letter.
Author Response File: Author Response.docx
