Origin Identification of Saposhnikovia divaricata by CNN Embedded with the Hierarchical Residual Connection Block

Round 1

Reviewer 1 Report

This manuscript contains interesting content. The research is complete and logical, but it appears that the authors have spent much time and effort collecting what appears to be a very large and potentially highly informative body of evidence. For that, I commend them. However, the manuscript in its current form does not fully present and interpret the results from this data set.

My comments are as followed:

1.      Some sentences are too long or contain grammar mistakes. It made the manuscript difficult to read. I only gave three here. However, the whole manuscript needs significant improvement. Additionally, throughout the text, there are different grammatical errors, the spelling of scientific names (it must be in italics), problems with the citation format, and the keywords being repeated in the title, among others.

2.     The introduction and discussion section are poorly described, the authors should include more details and comparisons on other similar studies in agriculture production worldwide. My suggestion is included in the introduction and discussion basics aspect associated with the relevance of use on CNN, integration of CNN into traditional classification tools, and new opportunities to apply novel tools such as CNN.

3.      The authors describe multiple networks in the introduction, but in materials and methods they evaluate others, it is not clear to me why? On the other hand, some basic theoretical concepts are given, however, there are no definitions or descriptions of them. I suggest that the authors present some comparisons or make a clearer description of it. I consider it necessary to explain the advantages and disadvantages.

4.      Experimental design and biological replicates are hard to follow from the M&M section. Please, insert a scheme to clarify the experimental design and sampling. More details are also needed to describe the analysis of the data so readers can ensure their appropriateness for the type of data presented. Additionally, please take attention to the detail needed to improve the overall quality of the manuscript including the small detail about the concept of CNN, data uses, statistics analysis, and software used.

5.      The sections are not well defined, I have lost myself in which ones correspond to materials and methods, results, and discussion.

6.      There is also general confusion as to which materials and methods and corresponding results are adequate to use by growers, the government, and the scientific community. More details are also needed to describe the analysis of the data so readers can ensure their appropriateness for the type of data presented. Additional attention to detail is needed to improve the overall quality of the manuscript including the small detail about the concept of CNN and its application to solve problems in agriculture.

7.      The discussion shows a limited understanding of the mechanisms involved in the development of the design and applicability of the CNN used in the study. Also, urge the authors to read all relevant scientific literature and re-write these sections. From my view, this part looks like a Conclusion rather than a Discussion. Authors need to cite some related kinds of literature and compare their data to others. Or explain your data in detail.

8.      Rewrite the Conclusion, try to be more specific and informative.

9.      Improve the Tables and Figures. Newlines in Tables made them hard to check. Figures are too hard to understand. Please check the main body. Make sure the order of description of them is the same as their real orders.

10.   Since this is pure computation work, I think that the availability of codes and data is essential to make possible the reproducibility of the study. It is not possible to check if the statistical analyses were performed correctly.

Minor revision

Title

The title is a little confusing. Moreover, it seems (and I could be wrong here) that the authors use the term “IResNet” in their title and throughout the manuscript as an advertising strategy to get readers' attention since CNN became mainstream in the last few years. Therefore, I would not say that their title described the main idea of their work.

Abstract

Try to be specific and write a paragraph more informative, because is very confused the aim. In addition, you can add more information based in data (statistical, among others).

 

Introduction

My suggestion is that aim, and hypothesis should be presented very clearly. In addition is necessary to improve, because currently, these are very wordy and confusing, in especially the role of the approach used.

 

Conclusion

The author should improve the conclusion and focus on the most important data of the study. The conclusions presented do not represent the importance of the research work.

 

Reference

Try to improve based on the before comments. In addition, review the correct format used by the Journal.

 

Figure and Tables

 

Poor resolution and need to be compressible standalone.

Author Response

尊敬的审稿人：

请参阅附件

Author Response File: Author Response.docx

Reviewer 2 Report

The authors presented promising work for classifying plant species (4 places of origin). The architecture they describe and implement is interesting and relevant, and the results are good enough. However, some concerns should be addressed.

- In general, some sentences and paragraphs are hard to read. I suggest improving the writing style. Moreover, upper-case is used after some commas+references, which needs to be corrected.

- It is necessary to improve the citation style.

- L27-28: Redundant phrasing. Please, rephrase.

- L57: VggNet -> VGGNet .

- Figure 1 (Caption): This is not enhancement but just data augmentation, a common technique to avoid overfitting.

- Is the “Ki()” notation usual? This is the first time I have read it.

- More images of the dataset should be provided in the manuscript—at least 2 per place of origin.

- A link to check the dataset is necessary.

- L278: Full-stop missed. Please, check the writing style of the whole manuscript. 

- L278: This phrase is difficult to read. The “and” is in the wrong place.

- Table 2: Training rounds? “Epochs” is a more usual term. 

- Table 7 should be integrated with Figure 7 and Figure 8.

- How would this classifier be used? Should the models be trained against other plants, which are also part of the same origin? 

- L318-320. This information should be provided while starting the paragraph.

- According to Figures 7 and 8, the network before the improvement is better. Is this right? I don’t think so.

- Images from mislabelled samples should be shown in order to make a visual/qualitative analysis of the image.

- A qualitative analysis using explainability techniques (e.g., Grad-CAM) should also be provided. This is a common practice in current papers, and it gives credibility to the research.

Author Response

Dear Reviewer:

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

The manuscript is considered to have proved the effectiveness of the model that the authors proposed. However, the effectiveness was only proved for the image classification but was not proved for the agricultural aspects. Further information may be required. 

1. The characteristics of Saposhinikovia divaricata for each origin should be provided. If the model can detect the difference that can be partly distinguished by human observarion, it would be very promising. However, if the model can calssify the origin but the reason is unknown, it is considered that some kind of overfitting has occurred. Even though the information is not scientific, the information may be useful on how human classify the origin. 

2. In order to avoid overfitting, the completely independant test data shold be prepared. 

3. The effect of imge enhancement is unknown. Figure 1 shows the enhancement fully changed the images. The enhancement might contribute the classification but it might be arbitrary. Procedures for image enhancement and objective evidence of their effectiveness should be presented. 

Author Response

Dear Reviewer:

Please see the attachment

Author Response File: Author Response.docx

Reviewer 4 Report

Thank you for submitting your work.

Please improve the reference section seems some references were copied and pasted from other works without adding the interspace among words.

  The introduction does not provide sufficient background. Must be improved.

all the cited references are relevant to the research. Nor applicable

Is the research design appropriate? must be improved

Are the methods adequately described? must be improved

are the results clearly presented? yes,

are conclusions supported by the results? Must be improved

 

Author Response

Dear Reviewer:

Please see the attachment

Author Response File: Author Response.docx

Reviewer 5 Report

The presented article devoted to the origin identification of Saposhnikovia divaricata. It is relevant for crop production. Recognition of different plants and their origins will automate the processes of plant collection, which are important for medicine. Nevertheless, it is worth noting a number of remarks to the article:

1. A more detailed description of the improved method should be given with a more detailed assessment of its effectiveness.

2. The article presents quantitative indicators in the form of tables and graphs, but it is not clear how they were obtained and are not substantiated in any way.

3. In order to analyze the recognition efficiency using various models, as well as to identify the distinctive features of images for which recognition was not successful, it is recommended to add an illustrative material containing images of Saposhnikovia divaricata with highlighted recognized areas.

4. The abbreviation TCM in the conclusion requires description.

5. Authors should also check the manuscript for typos and additional explanations.

Author Response

Dear Reviewer:

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

After the corrections, the authors have included the suggestions

Author Response

Dear Reviewers:
Thank you for reviewing the manuscript and giving your valuable comments.Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors have addressed all of the comments. I’m not very convinced by the use of data that can’t be checked by the reviewers, even in a read-only mode. Finally, there are still some issues to fix:

• L21: “( 86.68%)”. White-space missing.

 

• White spaces are missing before the citation brackets.

• Reference 25 misses white-spaces before and after.

• Table 1: “Image” in the header should be plural or be deleted.

• In section 3.2.2, you use “IIR stage”, but then in the Figure 3 caption, you use IIResStage. Is this right?

• Table 2: Typo -> “Imput Size”

• Table 2: Why do “optimizer” and “learning rate scheduler” (contrary to the other hyper-parameters) use lower-case?

• Please, read the paper carefully to avoid more typo issues and inconsistencies between the used terminology.

Author Response

Dear Reviewers:
Thank you for reviewing the manuscript and giving your valuable comments.Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Although the drastic color change due to image enhancement is still unfavorable for the reviewer, Table 4 provides enough information for the readers to judge the necessity of image enhancement. The reviewer acknowledge that the authors responded appropriately to the comments.

Author Response

Dear Reviewers:
Thank you for reviewing the manuscript and giving your valuable comments.Please see the attachment.

Author Response File: Author Response.docx

Reviewer 5 Report

The authors carefully paid attention to all comments.  Necessary changes have been made to the article in accordance with each of the comments.  The article can be accepted if there are no comments from other reviewers.

Author Response

Dear Reviewers:
Thank you for reviewing the manuscript and giving your valuable comments.Please see the attachment.

Author Response File: Author Response.docx