Axis Learning for Orientated Objects Detection in Aerial Images

Round 1

Reviewer 1 Report

The article presents a novel one-stage anchor-free method for detecting orientated objects. The subject is interesting and the presented method consists of an alternative to the current State-of-Art in terms of methodology, but it is not a breakthrough in terms of accuracy. The use of the OriCenterness increases the output quality.

1. In the Introduction is missing the main application foreseen for the method. Is it intended to be applied in UAVs or general aerial images? Used in Real-Time/Post-Processing?
   Despite the main idea can be absorbed, there are some sentences that can be misleading the real intended meaning due to some English inconsistency. For example: "This detector not only simplify the format of detection but also avoids elaborating hyperparameters and reduce computational complexity."
   As stated by the authors, the processing times comparison might not be fair, due to the difference in the hardware used.
2. In section 4.2, the extensive description of the values of the hyperparameters might be unnecessary since they are in Table 2.
   
   
3. As stated by the authors, the processing times comparison might not be fair as different devices are used. Regarding IENet, the processing times are very close, even running in a device with much less power than the one used by the authors in larger images. In the HRSC2016 dataset, the mAP is very similar (Table 7).
   If the desired application is the use in UAVs, some tests (in terms of inference times) shall be presented with lower power devices, such as the Jetson Nano.
4. The provided conclusions are too general and short. There is a lack of detail and exposition of the main values obtained. Future steps shall also be discussed.

The organization of the article must be revised. Especially from the Results section, the document becomes too confusing and really hard to read due to the distance between the text and the corresponding tables and/or images.

The authors are strongly advised to review the English-style of the document. There are a lot of misspellings, punctuation and grammar errors to fix.

Author Response

Dear Reviewers:

Thank you very much for your comments concerning our manuscript entitled “Axis Learning for Orientated Objects Detection in Aerial Images”.

We have revised the manuscript and please see the attachment for point-by-point response.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is an interesting paper with a clear presentation of the result. My only recommendation to the authors is to perform an oveall review of the redaction of the whole document. 

Author Response

Dear Reviewers:

Thank you very much for your comments concerning our manuscript entitled “Axis Learning for Orientated Objects Detection in Aerial Images”.

We have revised the manuscript and please see the attachment for point-by-point response.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors propose a new one-stage anchor-free method to detect orientated objects in per-pixel prediction fashion with less computational complexity. The manuscript is well written, and the topic is interesting. I want to suggest an expansion of the conclusion. Also, including the limitations of the methodology and plans to address those.

Author Response

Dear Reviewer:

Thank you very much for your comments concerning our manuscript entitled “Axis Learning for Orientated Objects Detection in Aerial Images”.

We have revised the manuscript and please see the attachment for point-by-point response.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

First I would like to thank you the authors for taking into consideration all the comments for the previous version.
The rearrangement of the structure made it much easier to read.

As a final note, when you say in Conclusion:
"... there are some limitations such as requirements for high quality label data, and the accuracy is not state of the art."
This is not completely true. Your method can be considered as a State-of-Art solution. My previous comment was that it is not a breakthrough in accuracy in the sense that the overall accuracy values do not show a clear improve for all situations. However, it is comparable with the existent methods.

Author Response

Dear Reviewer:

Thank you very much for your comments concerning our manuscript entitled “Axis Learning for Orientated Objects Detection in Aerial Images”.

We have revised the manuscript and please see the attachment for point-by-point response.

Author Response File: Author Response.docx