Multi-Scale Feature Interaction Network for Remote Sensing Change Detection

Round 1

Reviewer 1 Report

The manuscript attempts to propose a Multi-scale Feature Interaction Network for Remote Sensing Change Detection, it is interesting, the main problems are following

1）in line 23 and 24, ‘natural disaster assessment’ appears twice, pls check;

2) the english needed to be proved by a native english speakers;

3) Figure 3. is wrong;

4) please provided the program and data on the GitHub to be tested and verified by others,

Author Response

The authors are grateful to the anonymous reviewers for their kind help and constructive comments which helped improving the presentation of the paper.

Response to reviewer#1

Comment 1: in line 23 and 24, ‘natural disaster assessment’ appears twice, pls check.

Response: Thank you very much for your comments on this paper. The duplicate parts of this article have been deleted.

Comment 2: the English needed to be proved by a native English speakers.

Response: Thank you very much for your comments on this paper. We have invited native English teachers to sort out according to your request. This section has been highlighted in blue.

Comment 3:  Figure 3. is wrong.

Response: Thank you very much for your comments on this paper. I have reconstructed the flowchart for Figure 3, as detailed in the article Figure 3.

Comment 4:  please provided the program and data on the GitHub to be tested and verified by others.

Response: Thank you very much for your comments on this paper. Thank you very much for your suggestion. We are preparing to upload the paper after its official publication.

 Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper proposes a CD algorithm extracting multi-scale feature information through the backbone network in the coding stage. The research design is appropriate. The paper will be interesting for the readers, but it can be improved before it is accepted for publication. I recommend the authors to:

- state the limitations of their algorithm and plans for future work in the field at the end of the Abstract;

- review in-depth the existing algorithms;

- state the aims of the paper and describe its structure at the end of the Introduction;

- move innovation points of the article from the Introduction to the Conclusion section;

- describe the methodology in more detail and state why they use Multi-scale Feature Extraction Network, Feature Interaction Module, Detail Feature Guidance Module, and MLP Decoder;

- describe in more detail the LEVIR-CD data source.

- state the limitations of their model and plans for future work in the Conclusion

 

 

 

 

Author Response

The authors are grateful to the anonymous reviewers for their kind help and constructive comments which helped improving the presentation of the paper.

 

Response to reviewer#2

Comment 1: state the limitations of their algorithm and plans for future work in the field at the end of the Abstract.

Response: Thank you very much for your comments on this paper. We did not explain the limitations of the algorithm and future work plans in the abstract section of the previously submitted article, which was my oversight. Thank you very much for your reminder. In the latest submitted article (L16-L21), I have supplemented the limitations of the algorithm and future work plans; This section has been highlighted in red.

Comment 2：review in-depth the existing algorithms.

Response: Thank you very much for your comments on this paper. In the previously submitted manuscript, (L38-L50) has already explained the development of CD detection, as well as detailed the algorithms used at each stage, and also explained the limitations and difficulties of using different algorithms at different stages. Then, with the progress of technology and the accumulation of data, and based on the continuous efforts and innovative research of previous scholars, we finally proposed the current algorithm.

Comment 3: state the aims of the paper and describe its structure at the end of the Introduction.

Response: Thank you very much for your comments on this paper. At the end of the introduction section of this article, I have supplemented the description of the overall structure of the article (L76-L79). The rest of the paper is as follows: Section 2 introduces each network in the model. Section 3 analyzes the composition of the data set. Section 4 analyzes the performance of the model through experiments. Section 5 summarizes the whole paper and puts forward the future work.

Comment 4: move innovation points of the article from the Introduction to the Conclusion section.

Response: Thank you very much for your comments on this paper. Your suggestion is very good. I have described the innovative points in the conclusion section of this article according to your suggestion（L310-L316）.

Comment 5: describe the methodology in more detail and state why they use Multi-scale Feature Extraction Network, Feature Interaction Module, Detail Feature Guidance Module, and MLP Decoder.

Response: Thank you very much for your comments on this paper. In the final section of the introduction in this article, I added the detailed functions and advantages of using the Multi scale Feature Extraction Network, Feature Interaction Module, Detail Feature Guidance Module, and MLP Decoder(L76-L86).

Comment 6: describe in more detail the LEVIR-CD data source.

Response: Thank you very much for your comments on this paper. In the dataset description of the article, a detailed description of the source and category of LEVIR-CD data has been provided(L223-L229).

Comment 7: state the limitations of their model and plans for future work in the Conclusion.

Response: Thank you very much for your comments on this paper. The conclusion of this article has restated the limitations of the algorithm and future work plans(L317-L322).

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This article proposes a Multi-scale Feature Interaction Network (MFIN) model to detect changing regions in dual temporal remote sensing (RS) images. This method is very worth studying.

Extracting multi-level feature information from dual temporal images through the Multi-scale Feature Extraction Network (MFEN). Then, two sub modules, the Feature Interaction Module (FEM) and the Detail Feature Guidance Module (DFGM), are used to effectively interact with the feature information of the dual temporal image.

There are several small issues that need to be addressed in this article:

1.The author did not describe the existing algorithm in the introduction section, and the introduction section should supplement the existing change detection algorithm.

2. Some languages in the article are not written in strict accordance with English grammar, and grammar in the article needs to be corrected.

3. The limitations of the algorithm were not explained in the conclusion of the article, and the limitations and future work plans of the algorithm were explained in the conclusion.

4. The limitations of the algorithm are not described in the abstract section of the article. Please provide a description in the abstract section.

Author Response

The authors are grateful to the anonymous reviewers for their kind help and constructive comments which helped improving the presentation of the paper.

Response to reviewer#3

Comment 1: state the limitations of their model and plans for future work in the Conclusion.

Response: Thank you very much for your comments on this paper. In the previously submitted manuscript, (L38-L50) has already explained the development of CD detection, as well as detailed the algorithms used at each stage, and also explained the limitations and difficulties of using different algorithms at different stages. Then, with the progress of technology and the accumulation of data, and based on the continuous efforts and innovative research of previous scholars, we finally proposed the current algorithm.

Comment 2: Some languages in the article are not written in strict accordance with English grammar, and grammar in the article needs to be corrected.

Response: Thank you very much for your comments on this paper. We have invited native English teachers to sort out according to your request. This section has been highlighted in blue.

Comment 3: The limitations of the algorithm were not explained in the conclusion of the article, and the limitations and future work plans of the algorithm were explained in the conclusion.

Response: Thank you very much for your comments on this paper. The conclusion of this article has restated the limitations of the algorithm and future work plans(L317-L322).

Comment 4: The limitations of the algorithm are not described in the abstract section of the article. Please provide a description in the abstract section.

Response: Thank you very much for your comments on this paper. We did not explain the limitations of the algorithm and future work plans in the abstract section of the previously submitted article, which was my oversight. Thank you very much for your reminder. In the latest submitted article (L16-L21), I have supplemented the limitations of the algorithm and future work plans; This section has been highlighted in red.

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The paper can be accepted