Artificial Reef Detection Method for Multibeam Sonar Imagery Based on Convolutional Neural Networks
, Yanxiong Liu 1,2, Long Yang 1,2,*, Yikai Feng 1,2, Jisheng Ding 1,2 and Fengbiao Jiang 1Round 1
Reviewer 1 Report
1. This paper proposed an AR-Net framework to realize artificial reef detection and extraction in multibeam sonar images. In the reviewer’s opinion, the proposed framework does not focus on the characteristics of the multibeam sonar images, it can also be used for many other images, such as natural images, medical images. In other words, the proposed method lacks enough novelty to be published. The authors should concentrate more on the special properties of multibeam sonar images.
2. In equation 3, the coefficients of R, G, B are set to be 0.299, 0.587, 0.114. This needs further explanation, what is the principle of setting these coefficients?
3. Although the evaluation criteria of recall, F1 score, IOU of the proposed algorithm are the best, the precision of the proposed algorithm is not the highest. The phenomenon means that some true reefs have not been detected. And the undetected reefs cannot be recovered in the following process. How to solve this problem?
4. The computation complexity of the proposed algorithm should be evaluated.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors did excellent work for reef area semantic segmentation by AR-Net, a convolutional-based network. There are my comments and suggestion.
The authors mention their AR-Net obtained the best result in large-scale multibeam sonar images compared to other literature. Can the author construct an experiment with large-scale and small-scale data?
The authors analyze the model complexity and efficiency by only model #parameters. It is not sufficient to claim the model efficiency. Can the author provide the FLOPS of each model?
Since the dataset is small, I suggest constructing the k-fold cross-validation to include in this work.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Very nice, well written and well-presented article titled” “Artificial Reef Detection Method for Multibeam Sonar Imagery Based on Convolutional Neural Networks”
The experimental results demonstrate that the proposed method can achieve 86.86% F1-score and 76.74% intersection-over-union (IOU), and outperforms some state-of-the-art artificial reef detection methods. I appreciate the efforts made by authors.
The paper structure lags the research methodology, as it is not clear, and results are added one after other without establishing any comprehensive link.
For a study like this, it is important to separately add some recent literature Section,
I can see you have added the literature section, but I recommend presenting the comparison in tabular and comprehensive manner
Conclusion section can be improved by highlighting key findings, limitation of the research and recommendation for future studies. Its just like you are comparing results giving your perspective for future.
Methods used in the study are already compared well with the efficiency of one method over the other, but Explain why the current method was selected for the study, its importance and compare with traditional methods (add discussion on it) .
Also add risk factors in your study – risk associated with this domain – risk matrix. / False positive and True negative should be added.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
The authors studied artificial reef detection for multibeam sonar imagery based on CNN. The complexity of the model and the useful results provide a good novelty research area. The manuscript is well structured and the writing and language used are easy to comprehend. This will help future researchers to be able to use it. However, I have some minor comments as listed:
- It would be good to discuss some limitations in the conclusion.
- Please add the total number of parameters for your model.
- A "Related Work" section is completely missing, which highlights the most important techniques (at least for the chosen problem), and above all a comparison with the technique proposed in the paper. A final table summarizing the technical aspects of the various methods would be very useful.
- Make sure that the Conclusion briefly summarizes the results of the paper it should not repeat phrases from the Introduction.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The calculation details of the FLOPS of the six algorithms in Table 4 should be given.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
