Direction Estimation of Aerial Image Object Based on Neural Network
Round 1
Reviewer 1 Report
The article is within the scope of the journal. The topic described is interesting.
The article is well written and easy to read. It is well structured.
The results obtained are interesting and the experiment carried out is described.
Nevertheless:
a) A discussion section must be included in which the results obtained with similar works are compared and the progress of the work as well as the limitations are indicated.
b) The state of the art of the article should be extended.
c) In the conclusions section, the scientific contribution of the article must be made clearer, as well as a list of lines of future work.
Author Response
Please see the attachment
Author Response File: 
Author Response.docx
Reviewer 2 Report
The Article proposes a rotating object direction estimation method based on a neural network, which determines the unique direction of the object by predicting the direction vector.
The introduction provides almost sufficient background. It should be enhanced. The research design is appropriate, even if a flow chart could help better understand the procedure. The methods are adequately described but the results and the conclusions have to be enhanced.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 3 Report
In this paper a method of object detection and direction estimation at the same time is presented. An accuracy index for quantitatively evaluating the accuracy of angle estimation is also proposed. The novelty seems to be the estimation of direction cosin (cos(angle), sin(angle)) instead of direct angle. And the introduction of a normalization layer in order to respect the well known trigonometric constraint (cos(t)^2+sin(t)^2=1).
As a general comments, the problem covered in the present paper actually should be more strongly motivated. Authors need to give more reason for the importance of the problem they address in the article (obtaining orientation of an object).
The needs of using direction cosine (cos(t), sin(t)) instead o the angle is the true novelty of the paper. As a justification a preprint paper is cited ([4] in the paper references).
Also using normalization layer is very common in the literature; see for example:
Santurkar, S., Tsipras, D., Ilyas, A., & Madry, A. (2018). How does batch normalization help optimization?. In Advances in Neural Information Processing Systems (pp. 2483-2493).
Wu, Y., & He, K. (2018). Group normalization. In Proceedings of the European conference on computer vision (ECCV) (pp. 3-19).
row 51 - You say that "At present, there are many studies on object direction estimation." Please provide citation which definitely show the importance of the problem.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The paper can be accepted in current form
Author Response
Dear reviewers,
Thanks so much for all the help from you. We have made careful modifications on the manuscript.
Thank you for your consideration.
Sincerely yours,
Zhang Hongyun
