Detection of Mechanical Damage in Corn Seeds Using Hyperspectral Imaging and the ResNeSt_E Deep Learning Network

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Why choose PCA, FA, KPCA, and other advanced dimensionality reduction methods, but not compare them in the manuscript? It should be explained in section 2.5.

No specific details were provided for PCA and KPCA, such as how to determine the number of principal components and their impact on the experimental results. It should be explained in section 3.2.

Are the FA+KPCA and KPCA+FA experiments in Tables 3, 4, and 5 different from each other? If so, what are the specific differences? Please provide a detailed explanation in section 3.3.

Comments on the Quality of English Language

Moderate editing of English language required.

Author Response

Thank you very much for taking the time to review this manuscript. Please see the attachment for the detailed responses. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript should be published in Agriculture after the authors have undertaken the necessary minor revisions.  Specific queries and criticisms that need to be addressed are listed below.

1.      The authors should provide a more detailed description of factor analysis and how it is related to principal component analysis.

2.      The authors should provide literature references for principal component analysis, factor analysis, and kernel principal component analysis.

3.      A more detailed explanation on how PCA and FA are coupled together for feature extraction would benefit the reader.

4.      How did the authors go about dividing the data set into a training set and validation set?

5.      Why didn’t the authors compute specificity and sensitivity in addition to accuracy for characterization of model performance?

Author Response

Thank you very much for taking the time to review this manuscript. Please see the attachment for the detailed responses.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The study compares the classification accuracy of different models on hyperspectral images on corn kernels. In my opinion this study shows no significant effort other than comparison of different classifiers, and lacks in several important areas to be considered for acceptance. See the detailed comments in the attached file.

 

 

Comments for author File: Comments.pdf

Comments on the Quality of English Language

Sub-optimal

Author Response

Thank you very much for taking the time to review this manuscript. Please see the attachment for the detailed responses.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This paper analyzes three feature extraction and combination of methods, principal component analysis (PCA), kernel PCA (KPCA), and factor analysis (FA). The authors through dimensionality reduction results (FA+KPCA, KPCA+FA, and PCA+FA) combined a new dataset to improve the classification effect. Comparing the effects of six classification models, they proposed a ResNeSt_E network based on the ResNeSt and efficient multi-scale attention modules. The authors claim that ResNeSt_E network can detect both internal and surface cracks in corn seeds, making it suitable for mobile terminal applications.

While the authors have provided a comprehensive overview of existing methods and models, the results and discussion sections could benefit from a more in-depth exploration of the findings and enhance the overall quality of the paper.

Here some spcific comments and open questions regarding the manuscript.

- Please consider adding the following paper to the reference list, as it is relevant to the field of precision agriculture: https://doi.org/10.3390/s24020344

- Some typos where found as "R0" at line 148, and "1,000 nm" at line 154 at page 4. So it is recommended to carefully proof read the manuscript.

- line 122 pag 3 "intact (IN; 600 kernels), broken (BR; 552), internally cracked (IC; 720), and surface-cracked (SC; 624)", please specify if the values reported are the kernels.

- Table 1 is not referenced in the text, same for table 3.

-The authors should provide a more complete discussion of the tables presented in Section 3. Additional context and explanations would improve the clarity and understanding of the data presented.

-The discussion in Section 4 should address specific results of Section 3 (for instance tables and/or figures wherever possible) and the authors should compare the findings to relevant state-of-the-art research.

Comments on the Quality of English Language

Minor editing of English language required, with some typos and redundant wording to fix.

Author Response

Thank you very much for taking the time to review this manuscript. Please see the attachment for the detailed responses.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have still not considered the importance of the below comments:

1- Please clarify which feature extraction and edge-detection approach was used for this work and justify your approach. No analysis, no selection criteria, no results. Again, whichever approach was selected for the entire dataset, or separately for the train/val/test sets? The rationale behind using all those different methods is still unclear and confusing.

2- Fig 2- No details at all on the dimensions for any of the models, except extremely suboptimal quality images that convey no meaning whatsoever. For this comment, providing the core modules is fine, but all those final models must be associated with the corresponding details on model structure, input-output dimensions, hyperparameters, etc. Nothing is again mentioned.

3- All the figure qualities must be improved. Nothing is legible now.

4- The overall rigor in this study seems poor. Please try to substantiate your approach with details and evaluations.

Comments on the Quality of English Language

Please make sure to lucidly present your rationale and discuss your results.

Author Response

Thank you very much for taking the time to review this manuscript. Please see the attachment for the detailed responses. 

Author Response File: Author Response.pdf