A Generic Self-Supervised Learning (SSL) Framework for Representation Learning from Spectral–Spatial Features of Unlabeled Remote Sensing Imagery

Round 1

Reviewer 1 Report

1- p7, Line 266, Eq1 : Explain and justify the choice of the SoftMax function. And mention the advantages compared to other functions such as Sigmoid…

2- In related work, add other references such as :

10.1109/SSD.2019.8893202

3- Line 422 : Explain and justify the choice of the Hamming Loss function. Is it possible to use others like the Euclidean, or … What about its performance?

4- Section 5 Result and 6 Discussion:

a-      It’s better to add ROC and AUC curve and their interpretation.

b-       About the dataSet : you compare your result with others, but you don’t precise if others works used the same or not dataset. You must justify your choice.

c-      Is it relevant to test your method with another dataset?

Good english

Author Response

Dear Editor and reviewers,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments (Please see the attachment) (response to reviewers), (b) an updated manuscript with yellow highlighting indicating changes, and (c) a clean updated manuscript without highlights (PDF main document).

Author Response File: Author Response.pdf

Reviewer 2 Report

This study proposed an interesting SSL framework for remote sensing imagery. The study is interesting for readers and achieves a higher accuracy than similar works. 

Further comments are as follows:

Please briefly explain the subsections of 2, 5, and 5.1.

Please justify the setup in 4.1.3.

Please double-check the accuracy of equations and ensure that their parameters are explained well.

For further comments, please refer to the attachment.

Minor editings are needed.

Author Response

Dear Editor and reviewers,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments (Please see the attachment) (response to reviewers), (b) an updated manuscript with yellow highlighting indicating changes, and (c) a clean updated manuscript without highlights (PDF main document).

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper proposed a novel SSL framework that is capable of learning representation from both spectra-spatial information of unlabeled data. The proposed method performs better than exists method in two different downstream tasks.

Overall, the paper is looked good, and I have some minor suggestions.

1, Please highlight the best performance in Table 4.

2. Please add average score in Table 3.

3, Could you provide more details about regression task. and some datasets description for  p,k and Mg.

4. I'm interested in the proposed augmentation methods, could you provide some ablation study on different augmentations and show some effectiveness of them.

Author Response

Dear Editor and reviewers,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments (Please see the attachment) (response to reviewers), (b) an updated manuscript with yellow highlighting indicating changes, and (c) a clean updated manuscript without highlights (PDF main document).

Author Response File: Author Response.pdf

Reviewer 4 Report

Major revision The subject of A generic self-supervised learning (SSL) framework for representation learning from spectra-spatial feature of unlabeled remote sensing imagery, is interesting and can be suitable for the Remote sensing journal. However, before recommending the acceptance of this article, several improvements need to be made:

1. The results in the abstract section should be presented more concisely. Additionally, the purpose of the article should be clearly stated in the abstract, along with a brief explanation of the research method.

2. The introduction section should provide a stronger background for the research. This will help readers understand the context and significance of the study.

3. The innovation of the research should be highlighted in the introduction. This will help distinguish the article from previous studies and emphasize its contribution to the field.

4. The quality of the figures should be improved. Clear and visually appealing figures will enhance the understanding and presentation of the research findings.

5. The limitations of the study should be clearly stated in the discussion section. This will provide a balanced view of the research and help readers understand the potential weaknesses or constraints of the study.

6. The discussion or conclusion section should include suggestions for future studies. This will encourage further research and provide ideas for expanding on the current findings. Best regards

Author Response

Dear Editor and reviewers,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments (Please see the attachment) (response to reviewers), (b) an updated manuscript with yellow highlighting indicating changes, and (c) a clean updated manuscript without highlights (PDF main document).

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Please arrange (organize) the figure 10 size.

Author Response

Thanks for the comments. We have modified the size of figure 10 in the revised manuscript.

Reviewer 3 Report

The author almost solved all my concerns, and I recommend accepting the paper.

Author Response

Thanks for the comments.