Statistical Adaptation Loss Improved SMALL Sample Ship Detection Method Based on an Attention Mechanism and Data Enhancement

Round 1

Reviewer 1 Report

The paper proposed a method for using less data to obtain the same or similar detection results. The experimental results show that the proposed method is promising. However, several aspects need more details and I would like to share with my specific comments and suggestions below:

1. A deep literature review should be given in the paper. The related work part is rough.

2. In the section “The dataset used in this paper", more datasets used in this paper, such as OpenSARShip and HRSID, should be supplemented. The number of datasets should be clearly specified in the paper.

3. What is “Scale adaptation loss”. What are the advantages of the new LOSS algorithm.

4. For Table 1, the difference in accuracy between using 30% and 100% datasets should be analyzed.

5. Typo errors. For example:

--in P10L403: " Table 1 shows the comparison…" should be " Table 2 shows the comparison…".

6. Please seek professional editing service to improve the writing and readability of the manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1. The references are not in order in the manuscript. This needs to be revised. 

2. English grammar check and sentence structures need to be checked.

3. On line 121, there is a reference named [7.6]. This should be removed or corrected.

4. Related works on “Object detection method based on a deep convolutional neural network” could be enhanced.

5. For Figure 1, descriptions of subfigures (a) and (b) should be added to the figure caption.

6. On line 184, “Reference [28] KISANTAL M” part should be corrected.

7. Title “3 The dataset used in this paper” can be changed to a more informative and briefer one, and can be moved as a subtitle under the "Materials and Methods”.

8. In the paragraph starting from line 403, there are some explanations about the information from Table 2. This part should be revised.

9. There is no reference to Table 3 within the manuscript.

10. FPN and YOLOv5 models performed better when using 100% of the SSDD data. This needs to be explain and discussed more within the Discussion part. It seems like the proposed approach doesn’t perform better when using 100% of the data as mentioned in the text.

 

11. Similarly, the sentence “Our model performed well when using 100% SSDD dataset and slightly better than other models” in lines 451 and 452 also needs to be supported.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I have no other questions.

Author Response

Thank you for your work and help.

Reviewer 2 Report

My 6th comment was not realized as I was supposed to be. In line 200, the term of reference should be erased and the sentence should be started as Kisantal [28] proposed…

The subtitle of “4.1. The dataset used in this paper” should be revised to “4.1. The dataset used in the study”. Because this study s not a paper yet.

I couldn't find the correction of my eighth comment, which was “In the paragraph starting from line 403 -original version- there are some explanations about the information from Table 2. This part should be revised. If there is any revision please indicate it as red in the revised version of the manuscript.

The reference to Table 3 within the manuscript should be given in the previous chapters if possible in the paragraph coming earlier than Table 3.

Model performance should be clarified in the Discussion section. The given explanation is not sufficient.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf