Center-Ness and Repulsion: Constraints to Improve Remote Sensing Object Detection via RepPoints

Round 1

Reviewer 1 Report

In this manuscript, authors proposed an object detection for remote sensing based on Center-ness and Repulsion, I have some question and suggestion:

1. The authors have done a lot of comparisons; however, why compare with YoloV3 instead of the latest Yolo, maybe authors have to explain the reasons.

 

2. The authors provide abundant experimental results, to improve the completeness and quality of the manuscript, I suggest the authors should provide some failure cases of the proposed method and some analysis.

 

3. In section 4, the authors provided ablation study; however, I wonder which sub section is ablation study for the proposed oriented repulsion regression loss. Because I wonder the performance of the proposed method using other loss instead of the proposed oriented repulsion loss, did I miss it ?

 

4. In section 4.5, authors discussed about limitations of the proposed method. Normally, small object detection is a issue, is small object detection a limitation for the proposed method?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a shape-adaptive repulsion constraint on point representation to capture geometric information of densely distributed remote sensing objects with arbitrary orientations. The topic is interesting and one of the hot ones in the field of remote sensing image processing. The paper is well-organized and written. The literature review is also sufficient, technical issues are all correct. Experimental results are convincing. This review only has the following issues.

(1) The contributions are suggested to be enhanced clearly.

(2) Some strongly related work should be cited and discussed in the discussion section. E.g., 10.1109/JSTARS.2022.3175191 and 10.1109/JSTARS.2022.3214889.

(3) For the parameters alpha and beta in (10), how to determine them in a real application?

(4) If possible, the algorithm environment with step-by-step items is suggested to be summarized, so that readers will know the details of the method clearly.

(5)The running time as well as the parameter size is suggested to be reported.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

After careful reviewing, I believe authors have addressed all my questions, and I have no other extended questions.