CF2PN: A Cross-Scale Feature Fusion Pyramid Network Based Remote Sensing Target Detection

Round 1

Reviewer 1 Report

Some abbreviations are not described in the text.

The related work must be presented, and the discussion of the results should be compared with the related work.

The methods and results are well explained.

The conclusion must refer to the future work.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The topic of research is current and it touch the problems of modern society and has potential for further development. The readership is good.
For the sake of completeness additional experiments are needed, because nowadays many current research are conducted on similar problem about remote sensing target detection. Please see for example papers published in this year 2020 and making research on the same DIOR dataset:

https://www.mdpi.com/2072-4292/12/1/143/htm

https://www.sciencedirect.com/science/article/abs/pii/S0924271619302825?via%3Dihub

The results obtained by the Authors are not satisfactory. Authors should working on the performance and accuracy of their results. Moreover, Authors should try any other dataset, maybe satellite images. There are many open-dataset like from Copernicus program made by ESA, NASA or HDEOS program from China. Please, also consider, what their proposition of algorithm do if the image will be hazy?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This work presents a Cross-Scale Feature Fusion Pyramid Network for target detection in remote sensing applications. Some of the sections in the manuscript are interesting but overall it requires major improvements in line with the following comments. 

1) Mention the name of the datasets in the abstract section also that you used in your experiments. Also, mention the improvements that you have made as compared to previous methods. 

2) Make a separate section for the related work, discuss the pros and cons of all the previous methods and also summarize them in the comparison table that should give a short description of "one or two lines" and strengths and weaknesses of the previous methods as well as proposed method. 

3) How your proposed network is different from the previous network [32]. Explain with the help of tables and also mention the total number of parameters for both. 

4) Represent the results of your "proposed method" through the confusion matrics for each class and precision-recall curves. 

5) Also evaluate your model using TGRS-HRRSD (High-resolution Remote Sensing Detection) dataset. 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Author’s responses for comments and extension of research are satisfactory.

The paper is ready for publishing in the current form after minor correction: scale and named of axes in figure 12 and 14 are unreadable.

In addition, the editorial part of the work should be corrected in accordance with the rules of the journal. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Thank you for addressing all of my comments. Paper can be accepted in the present form. 

Author Response

Thank you for your approval!