Instance Segmentation of Underwater Images by Using Deep Learning

Round 1

Reviewer 1 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

The authors improved the manuscript.

Reviewer 2 Report (Previous Reviewer 3)

Comments and Suggestions for Authors

I would like to thank the authors for correcting the article in according to my comments. Now the paper sounds better and I can recommend one to a further processing.

Reviewer 3 Report (Previous Reviewer 4)

Comments and Suggestions for Authors

no new comments

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The methodology is thorough and well-structured, addressing the unique challenges of underwater imaging. The use of image rotation, flipping, and Generative Adversarial Networks (GANs) for data set expansion is innovative. The detailed explanation of image preprocessing methods, including various enhancement algorithms, adds significant value. 

- provide a more in-depth analysis of segmentation results.

- expand the conclusion section.

Comments for author File: Comments.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors presented the results of using deep learning to improve instance segmentation and target recognition for underwater images.

The article is interesting and could be considered for publication in Electronics after making the following amendments.

Figure 1 could be made clearer by improving the quality of the images (e.g., larger size and higher resolution). It is also unclear what the dotted box in the figure represents. Figure 1b could be improved by explaining in more details the different stages.

It is unclear what language/software was used for coding and what type of computer was employed. This should be specified in the Methodology section.

Line 166: Sentence “the research task … image segmentation” is redundant and should be deleted

Line 169: Sentence “in other words, … be labelled” is redundant and should be deleted

Line 192: The full stop between “images” and “in this paper” should be deleted

Please avoid to repeat “in this paper” as it is an unnecessary repetition

Figure 6 is too large which makes it difficult to understand it. It would be better to split it into three separate figures

Line 395: Sentence “it is necessary … human subjectivity” is unclear and should be rephrased

The results presented in Table 2 should be explained more clearly to convince the reader that WAC images are better. Furthermore, WAC algorithm does not work well for images 1,3,4, and 5 and that makes it unclear why the authors state that WAC works best

Figure 7 needs to be improved as a the moment runs over two pages and the quality of Figures 7c and 7d is not good enough (labels are too small and resolution is poor)

Line 520: The sentence “the experiment … Figure 5-7” is redundant and should be deleted

Lines 521-526: This information should be moved to the Methodology section

Lines 532-538: This information is redundant and should be deleted

The captions of Figures 9-13 are unclear as the only difference is the number in bracket at the end of the caption. A more comprehensive explanation of what the image represents should be provided

Conclusion section does not summarise sufficiently well the key findings

Comments on the Quality of English Language

Quality of English needs to be improved as there are many repetitions and typos

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript raises a problem of processing of underwater image. This issue is interesting because very often underwater images are characterized by a color shift, a blur and a poor contrast. In this paper a method improving such defects of images are proposed. This method is based on a deep learning.

The paper is good organized. Authors conducted a literature review, in my opinion the references were good selected.

 The article should be improved before being published:

1. All used parameters in equations have to be explained in the text (e.g. mG, mR, mB, how this parameters are calculated? Equation (5));

2. Why values greater than 255 are set to 255 and values less than 0 are set to 0? It has to be explain. Did authors into account a possibility of normalizing obtained values to range [0,255]? In my opinion such approach could give different results, maybe better.

3. Equation (12), What values does I(x) take, greater or less than 1? It is important from view point of equation (17).

4. Equation (7). Authors should mention about a source of values of matrix.

5. Equation (9). What are the values of Xn, Yn i Zn? Are these values constant?

6. The Conclusions in this article, could be augmented majored in terms of the effectiveness of the proposed method and verified in experimental results.

7. Conclusions - please to develop the sentence "In future work ...".

Reviewer 4 Report

Comments and Suggestions for Authors

1- Comment on the percentages in figures 09 to 13

2- Because in some starfish it gave a percentage of 53%, well below the others

??

3- Comment on how these percentages can be improved?

4- Suggestions for future work