Augmented EMTCNN: A Fast and Accurate Facial Landmark Detection Network^†

Round 1

Reviewer 1 Report

Overall opinion:

The manuscript focuses on a relevant topic inside the applied sciences. The new CNN based method has a great impact on face recognition and other optical data processing tasks.

In my opinion, the manuscript contains an adequate number of relevant references.

After reading the abstract, it was checked by myself, that the actual version of this manuscript is a well-extended version of the former proceeding content, and the amount of the new content in the manuscript is more than the earlier published and presented version. 

In the introduction, the background of the research was introduced clearly.

The methodology and the used materials are clear, exact.

The aims and achieved results were great presented in the manuscript.

I suggest paying attention to the language, and the authors write their words in the manuscript in a uniform passive voice.

Remarks related to the Funding section:

There is a small contradiction between the statement in the authors' contribution ( E.H. proposed the topic and obtained the funding) and the next sentence (This research received no external funding). It should be fixed in the manuscript, that the research has received external or any support or not.

Congratulate, great work, great results.

Author Response

Thank you very much for your positive comments. We have studied comments carefully and tried our best to make corrections, which we hope to meet with approval. All the changes are highlighted in yellow in the manuscript. A detail response can be found in the attached file.

Author Response File: Author Response.pdf

Reviewer 2 Report

Overall, this is a very well written and professionally produced paper. It has interesting, and well carried out, work, which is described well. 

The "Conclusions" section might be further improved by briefly summarising this work's place in the literature, with some ideas for future work following from that. 

98: Sentence starting "Based on....." could be better written. Also, unnecessary use of "the" between "by" and "Vahid". 

173: Repeated phrase: "Dilated convolution....".

Author Response

Thank you very much for your positive comments. We have studied comments carefully and tried our best to make corrections, which we hope to meet with approval. All the changes are highlighted in red in the manuscript. A detail response can be found in the attached file.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors propose a framework for facial landmark detection by comparing different neural networks and techniques. Since the idea is interesting, there are major points that need to be addressed to improve the work:

- There still contains grammatical issues and typos in this manuscript. The authors should re-check and revise carefully in the next version. Also, some of previous works should be mentioned in the past tense.

- The authors did not show any measured metrics in their model (such as accuracy, AUC, ...). For most deep learning paper, they should mention them as a reference to convince that they can have a good model statistically. Now they only show some figures on their predicted examples, I think it is not enough.

- The authors should report the performance results on different dataset separately.

- CNN or deep learning is common and it has been used in a lot of works in biomedical fields such as PMID:28643394 and PMID:31319963. Therefore, the authors are suggested to add more references in this section.

- It is necessary to have an independent test on their model.

- How to select the optimal hyperparameters of the models?

- Which evaluation method that the authors used in this study?

- It is very important to compare their performance results with different published works, since it is a common problem and there are a lot of published works on it. When comparing, please make sure that all works use the same dataset and evaluation method.

- The source codes and final model should be released for reproducing their results.

Author Response

Thank you very much for your positive comments. We have studied comments carefully and tried our best to make corrections, which we hope to meet with approval. All the changes are highlighted in yellow in the manuscript. A detail response can be found in the attached file.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

My previous comments have been addressed satisfactorily.