Research on the Wild Mushroom Recognition Method Based on Transformer and the Multi-Scale Feature Fusion Compact Bilinear Neural Network
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors1. Author says “After data augmentation, the dataset increased to 297 140,552 images, which were subsequently divided into training and testing sets in an 8:2 ratio. The training set contains 112,406 images, while the testing set comprises 28,146 images.” Check the correctness of the data.
2. Details are required for augmentation techniques such as rotation, scaling, and vertical flipping.
3. What is the source of data set ; It includes 100 species of wild mushrooms, with a total of 77,400 images.?
4. It is required to give details for image size from different data set.
5. Author discussed about classification of the mushroom. However, the age of the mushroom also plays a role? Why authors have not considered this?
6. What is the difference between micro and macro average recall? Why it is required for analysis?
7. Mention the proposed model in Table 1. Comparison of classification performance of various models. Indicate Technology name along with proposed.
8. Why large difference in ShuffleNet_V2 as compared to that of other models as per Table 1. Comparison of classification performance of various models. Better author can remove this model.
9. Not able to visualize different parameters in Figure 6. Quantitative evaluation of classification performance of each model. Propose different method to see all variables in this figure.
1. Similarly in Table 2. Comparison of overall average indicators of each model (Similar to table 1 representation; Which one is the proposed model?)
1. Readability and visibility is very poor in confusion matrix figure (Figure 7. Result confusion matrix of this model)
1. What is the tool has been used to generate thermal images as shown in Figure 8. Visualization of result thermodynamic diagram?
Comments on the Quality of English LanguageMinor editing of English language required.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors. Though the work is largely well-written, the article still needs significant improvement in the abstract, introduction and methodology and discussion sections. Considering my observations as follows, I suggest major revisions before considering it for publication.
1. In the abstract the author much more emphasized given on the methods rather results and concussion with policy recommendations. Please rewrite the abstract.
2. Is it necessary to discuss elaborately about MTC-BCNN Network Structure in the manuscript (section 2)? I think it is not so important. In the present form, section 2 would create redundancy in the manuscript. The author could discuss its structure either in the methodology or in the introduction section very briefly, not elaborately.
3. I suggest to rewrite the introduction with 3 paragraphs, highlighting the basic content of the research field in the first paragraph. Then review the research progress of the literature in the second paragraph, and in the third paragraph, analyze the limitations of past research and clarify the innovation of your own research.
4. It is very much important to describe the architecture of the proposed method in the methodology section very briefly.
5. The authors proposed a wild mushroom recognition method based on a compact bilinear neural network with Transformer and multi-scale feature fusion. There should be a discussion of the advantages, disadvantages, and limitations of the proposed methods in the discussion section.
Comments on the Quality of English Language
Moderate editing of English language required.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper focuses on the accurate identification of wild mushrooms, especially the requirement of identifying toxic mushrooms to avoid poisoning. The paper conducts research on wild mushroom identification method based on a compact bilinear neural network with Transformer and multi-scale feature fusion. Adopting a dual stream structure, two different feature extractors are used to extract and fuse features to capture more comprehensive image information. Designed a hierarchical perception transformation module to learn local features and global contextual information. The bottleneck attention modules and efficient multi-scale attention modules are embedded in the dual stream structure to capture multi-scale feature information. The results reported in the paper indicate that the accuracy of the proposed method reached 98.03%. The paper show certain innovations in the construction of recognition models and module customization.
There are still some shortcomings in the paper:
1) The paper lacks consideration for the practical application scenarios, deployment, and other aspects of the algorithm, resulting in insufficient practicality of the method.
To achieve non-expert identification of toxic mushrooms, convenience is an important aspect, and deployment on mobile devices seems practical. The model and algorithm in this article are too complex and lack the characteristics and corresponding tailoring solutions for mobile deployment.
2) The paper lacks data on the effectiveness and computation time of the algorithm. In terms of statistical analysis of experimental data in the paper, there is a lack of data such as computational complexity, running time, storage capacity, and parameter quantity during training and recognition.
) The paper lacks a unique image collection, as well as recognition types, scene descriptions, or limitations, resulting in insufficient research features. The paper lacks explanation on the classification and identification of toxic mushrooms. Which toxic mushrooms are difficult to distinguish, what are the difficulties, or which toxic mushrooms are easily confused with non-toxic mushrooms?
4) There are some shortcomings in the writing and formatting of the paper.
There are some vocabulary errors, some words are inconsistent in the paper.
Most of the time, Wild Mushroom appears multiple times , but Wild Fungi appears several times in Chapter 4.
In P10, L338, what is the meaning of iiith?
In the Conclusions chapter, there are multiple instances where words are split across lines.
Comments on the Quality of English LanguageModerate editing of English language required.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsAuthors have addressed all the points.
Author Response
Thank you very much for your professional comments and guidance on our article, which helped us improve its quality.
Reviewer 2 Report
Comments and Suggestions for AuthorsThough the authors replied to my previous comments, they did not correct my previous comments as it remained the same as before. For example, the abstract remains the same. Section 2 remains the same; no change has been made. I suggested to rewrite the abstract, focusing more on results, conclusion, and policy recommendation, but it was unchanged. I suggested to move Section 2 either in the introduction or in the methodology section, but it remained the same. I recommended to discuss the proposed model architecture very briefly in the methodology section, not elaborately, but I could not find it. Therefore, I cannot recommend for publication this research in its present form.
Comments on the Quality of English LanguageModerate editing of English language required.
Author Response
Thank you very much for your comment.Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe previously mentioned problems have already been addressed appropriately, and there are no other questions.
Author Response
Thank you very much for your professional comments and guidance on our article, which helped us improve its quality.
