Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Based on my detailed review of your manuscript, here are seven comments to help improve its quality for the publication:

1. While the introduction explains the importance of graphene ore detection, the specific gaps in existing literature are not explicitly detailed. Clearly stating what previous studies lack and how your method addresses those shortcomings would strengthen your rationale.
2. Although the manuscript presents the DW-YOLOv8 model clearly, it does not sufficiently justify the selection of the UniRepLKNetBlock and WIoU modules over other alternatives. Discussing these choices explicitly—ideally with references or preliminary tests—would enhance the methodological rigor.
3. In your research, YOLOv8 was employed, which is meaningful and shows solid performance. However, it is recommended that you cite the following two recent studies that utilized a more advanced YOLO version (YOLOv10) and discuss the limitations of YOLOv8 compared to YOLOv10: 1) Automated non-PPE detection on construction sites using YOLOv10 and transformer architectures for surveillance and body-worn cameras with benchmark datasets. 2) Heavy equipment detection on construction sites using You Only Look Once version 10 (YOLOv10) with transformer architectures."
4. Figures 9 and 11 (pages 11 and 13) lack clarity due to resolution and labeling issues. Enhancing these visualizations with higher resolution and clear axis labels would significantly improve readability and the manuscript’s professional presentation.
5. While the dataset collection and labeling procedure is described, more information about class distribution, variability in ore appearance, and potential biases would improve transparency and facilitate reproducibility (page 9-10).
6. The manuscript provides a good comparative analysis of models, but it would benefit from deeper interpretation and practical implications of the results (pages 13-15). Explicitly linking the model's improved performance to operational benefits in real-world production scenarios would add significant value.
7. Throughout the manuscript, minor grammatical issues, punctuation errors, and inconsistent terminology reduce clarity (e.g., abstract and sections 2-4). A thorough proofreading and language editing are recommended to ensure technical accuracy and readability.

Author Response

Comments 1:[While the introduction explains the importance of graphene ore detection, the specific gaps in existing literature are not explicitly detailed. Clearly stating what previous studies lack and how your method addresses those shortcomings would strengthen your rationale.]

Response 1:[Thank you for your insightful review—we have revised lines 105–108 to explicitly delineate research gaps in prior graphite ore grading studies and our corresponding methodological innovations.]

Comments 2:[Although the manuscript presents the DW-YOLOv8 model clearly, it does not sufficiently justify the selection of the UniRepLKNetBlock and WIoU modules over other alternatives. Discussing these choices explicitly—ideally with references or preliminary tests—would enhance the methodological rigor.]

Response 2:[We sincerely appreciate your highly pertinent and professionally rigorous critique. While supplementary experiments in Heilongjiang Province (China) would strengthen our validation, current funding constraints preclude immediate field research. We acknowledge this limitation and will prioritize data collection during future fieldwork opportunities in the region. In the interim, we have comprehensively revised the Introduction (lines 150–154) and Related Work section (lines 105–115). ]

Comments 3:[In your research, YOLOv8 was employed, which is meaningful and shows solid performance. However, it is recommended that you cite the following two recent studies that utilized a more advanced YOLO version (YOLOv10) and discuss the limitations of YOLOv8 compared to YOLOv10: 1) Automated non-PPE detection on construction sites using YOLOv10 and transformer architectures for surveillance and body-worn cameras with benchmark datasets. 2) Heavy equipment detection on construction sites using You Only Look Once version 10 (YOLOv10) with transformer architectures."]

Response 3:[Revisions have been made to the citation entries in lines 78–80 to ensure compliance with academic referencing standards.]

Comments 4:[Figures 9 and 11 (pages 11 and 13) lack clarity due to resolution and labeling issues. Enhancing these visualizations with higher resolution and clear axis labels would significantly improve readability and the manuscript’s professional presentation.]

Response 4:[We sincerely appreciate your meticulous observation. The revised figures with significantly enhanced clarity have been re-uploaded, specifically addressing the issues noted at lines 375 and 416.]

Comments 5:[While the dataset collection and labeling procedure is described, more information about class distribution, variability in ore appearance, and potential biases would improve transparency and facilitate reproducibility (page 9-10).]

Response 5 :[The dataset encompasses multi-perspective imaging of ore samples. However, confidentiality agreements with industrial partners preclude further disclosure of dataset transparency metrics, potentially constraining commercial deployment scalability.]

Comments 6:[The manuscript provides a good comparative analysis of models, but it would benefit from deeper interpretation and practical implications of the results (pages 13-15). Explicitly linking the model's improved performance to operational benefits in real-world production scenarios would add significant value.]

Response 6:[This model focuses on enhancing industrial efficiency in graphite ore processing through AI-driven digital transformation of measurement workflows. As one component of a larger interdisciplinary project, our laboratory collaborates with specialized teams handling image processing and system deployment. Deeper integration with operational cost-benefit analysis falls under ongoing research by collaborating researchers. Per institutional agreements and pending commercialization milestones, detailed theoretical frameworks involving team members' contributions cannot be disclosed at this stage. We appreciate your understanding regarding these collaborative research constraints.]

Comments 7:[Throughout the manuscript, minor grammatical issues, punctuation errors, and inconsistent terminology reduce clarity (e.g., abstract and sections 2-4). A thorough proofreading and language editing are recommended to ensure technical accuracy and readability.]

Response 7:[Thank you for your meticulous review—we fully endorse your technical recommendations and have systematically revised domain-specific terminology (e.g., replacing "graphene" with "graphite ore" throughout) to enhance precision and align with mining industry standards.]

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The paper presents a promising approach, DW-YOLO: A model for identifying the surface characteristics of graphene ore and distinguishing taste.
Authors are suggested to clarify the following comments:
1. This paper was created based on an artificial intelligence tool (e.g., Gemini). Please rewrite and fix all the references properly.   
e.g. deep[14] learning [13]
2. Please write the contribution point at the end of the introduction. Moreover, adda  paper structure at the end of the introduction 

3. Please include the related work section. (in section 2)
4. Equation 1 format is wrong
5. Figure 4: Overall structure of PAFPN (outside of the page)
6. Equations 2 and 3. Please follow the proper standard
7. Figure 6, please include a clear image.
8. What is the necessity of equation 6,7
9. Figure 9. Model Training Results (part of fig missing)  
10. Table 3-3. Ablation Study Results of the Overall Modules (please maintain proper structure)
11. Reference format is not correct; each reference contains [j], Reference [5] font size is not correct.    

Comments on the Quality of English Language

The quality of the English Language of this paper requires improvement. 

Author Response

Comments 1:[This paper was created based on an artificial intelligence tool (e.g., Gemini). Please rewrite and fix all the references properly.   e.g. deep[14] learning [13]]

Response 1:[Thank you for your thorough review of our manuscript. We highly value your insightful suggestions and have incorporated the recommended revisions in lines 52-55. Additionally, we have updated the relevant citations as suggested]

Comments 2:[Please write the contribution point at the end of the introduction. Moreover, adda  paper structure at the end of the introduction ]

Response 2:[Thank you for your valuable suggestions—we fully concur with the technical validity of your comments. Revisions have been incorporated in lines 150–154 as recommended.]

Comments 3:[Please include the related work section. (in section 2]

Response 3:[Thank you for your feedback—we have incorporated supplemental content in lines 105–115 per your recommendation.]

Comments 4:[Equation 1 format is wrong]

Response 4:[Thank you for your guidance—we have corrected the formatting of Equation (1) at line 250 per journal standards.]

Comments 5:[Figure 4: Overall structure of PAFPN (outside of the page)]

Response 5:[Thank you for your suggestion—we have optimized the dimensions of Fig. PAFPN at line 245 per visualization standards.]

Comments 6 :[Equations 2 and 3. Please follow the proper standard]

Response 6:[Thank you for your feedback—we have corrected the formatting of Equations (2) and (3) at line 275 per journal specifications.]

Comments 7:[Figure 6, please include a clear image.]

Response 7:[Thank you for your note—we have submitted revised versions of the figures with enhanced resolution per journal requirements.]

Comments 8:[What is the necessity of equation 6,7]

Response 8:[Thank you for your comments—we would like to clarify that Equation (6.7) explicitly defines the computational methodology for Average Precision (AP), a core evaluation metric in our framework. The AP metric is rigorously applied throughout the experimental results.]

Comments 9:[Figure 9. Model Training Results (part of fig missing) ]

Response 9:[Thank you for your guidance—we have implemented the revisions at line 367 per your recommendation.]

Comments 10:[Table 3-3. Ablation Study Results of the Overall Modules (please maintain proper structure)]

Response 10:[Thank you for your guidance—we have implemented the revisions at line 415 per your recommendation.]

Comments 11:[Reference format is not correct; each reference contains [j], Reference [5] font size is not correct.  ]

Response 11:[Thank you for your guidance—we have standardized all citations per journal specifications: references now appear immediately preceding punctuation (e.g., [5]), with font size adjusted at line 489.]

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript presents an apparatus and a method for analyzing graphite ore samples through surface imaging to estimate ore grade. The core idea, utilizing image-based analysis for mineral characterization, is potentially valuable and aligns well with current trends in automated mineralogy. However, the current version of the work suffers from important drawbacks in organization, clarity, technical rigor, and completeness. These issues prevent a full understanding of the scientific contribution, hinder the assessment of the methodology’s validity, and limit the possibility of reproducing the results.

One of the main concerns lies in the lack of precision in both the research objectives and the terminology used. The title and abstract suggest a focus on graphene characteristics and even use the term “taste,” which is scientifically ambiguous and undefined in this context. In contrast, the dataset and methodology are based on graphite ore samples, not graphene. This inconsistency creates confusion about the actual scope of the study. Furthermore, it remains unclear whether the aim is to classify graphite grades, detect impurities, or correlate visual features with chemical composition. The title and abstract should accurately reflect the focus on graphite analysis, non-standard terms should be defined or removed, and the research objective should be explicitly stated in the introduction.

The manuscript also lacks a proper Related Work section, making it impossible to evaluate the novelty or advancement of the approach compared to prior research in image-based mineral classification, automated ore grading, and the application of deep learning, particularly object detection models such as YOLO, in geosciences. Similarly, no Theoretical Background is provided to introduce the essential concepts of graphite ore characterization, image processing techniques, or the fundamentals of the algorithms used. Without these sections, the study lacks the contextual foundation that would allow readers to appreciate its contribution.

The methodology section is incomplete and insufficiently rigorous. Essential details such as dataset preparation, preprocessing steps, modifications to the YOLOv8 architecture, training procedures, hyperparameter choices, and evaluation metrics are either missing or superficially described. Mathematical formulations, where present, are poorly formatted, and equations lack a proper definition of variables. A step-by-step description of the entire pipeline, from image acquisition to model evaluation, as well as correctly typeset equations and clearly defined terms, would substantially improve this section.

Another limitation is the lack of robust experimental validation. The performance evaluation and ablation studies do not include meaningful comparisons with baseline models or state-of-the-art methods. Without such benchmarks, it is difficult to determine whether the proposed modifications to YOLOv8 offer real improvements. A comparative analysis with standard approaches would strengthen the claims.

The description of the experimental apparatus is also insufficient. Although Figure 7 illustrates the setup, there is no detailed account of its design or specifications. Important parameters such as lighting conditions, camera resolution, distance from the sample, and calibration procedures are omitted, yet these factors directly affect image quality and, consequently, model performance. Providing comprehensive specifications, diagrams, and a clear operational workflow would be essential for reproducibility.

In addition, the manuscript does not indicate whether the dataset or source code will be made publicly available. Transparency and reproducibility are essential in modern scientific research and sharing both would allow others to build upon this work. Details about the dataset, such as number of samples, resolution, labeling protocol, class distribution, and data splits, should be included, along with links to public repositories hosting the data and source code.

Finally, the overall presentation of the figures requires improvement. Several images are low-quality, cropped, blurry, or contain incomplete and unreadable text. This undermines the clarity of the evidence presented. All figures should be revised to meet high-resolution standards and comply with the journal’s formatting requirements.

Therefore, I do not recommend the publication of this work in its present form.

Author Response

Comments 1:[surface imaging to estimate ore grade. The core idea, utilizing image-based analysis for mineral characterization, is potentially valuable and aligns well with current trends in automated mineralogy. However, the current version of the work suffers from important drawbacks in organization, clarity, technical rigor, and completeness. These issues prevent a full understanding of the scientific contribution, hinder the assessment of the methodology’s validity, and limit the possibility of reproducing the results.]

Response 1:[

Thank you for your expert guidance—we have fully endorsed your critique and implemented the following revisions:

• Restructured Introduction (lines 105–115): Enhanced Related Work with explicit gap analysis against prior graphite ore grading studies;
• Added structural framework (lines 150–154): Clarified paper organization to unify the industrial deployment objective;

]

Comments 2:[One of the main concerns lies in the lack of precision in both the research objectives and the terminology used. The title and abstract suggest a focus on graphene characteristics and even use the term “taste,” which is scientifically ambiguous and undefined in this context. In contrast, the dataset and methodology are based on graphite ore samples, not graphene. This inconsistency creates confusion about the actual scope of the study. Furthermore, it remains unclear whether the aim is to classify graphite grades, detect impurities, or correlate visual features with chemical composition. The title and abstract should accurately reflect the focus on graphite analysis, non-standard terms should be defined or removed, and the research objective should be explicitly stated in the introduction.]

Response 2:[We sincerely appreciate your critical observation regarding the erroneous use of "taste" in the title—we have replaced it with the domain-standard term "grade" throughout the manuscript. Additionally, we have:

• Corrected all references from graphene to graphite ore (e.g., Section 2.1, lines 48–52);
• Eliminated non-standard nomenclature (e.g., "quality level" → "categorical grading");
• Restructured the Introduction (lines 105–115) to explicitly:
  (i) Define the research objective: Industrial-grade classification via surface texture analysis;
  (ii) Critique prior limitations: Lack of integration between visual features and chemical assay validation;
  (iii) Clarify methodological novelty: DW-YOLO’s dual-path validation against lab-measured carbon content.
  This study establishes a direct correlation between ore surface characteristics (Fig. 3) and chemically verified grade levels (Table 2), demonstrating the model’s industrial applicability through reproducible accuracy metrics (94.7% mAP@0.5).]

Comments 3:[The methodology section is incomplete and insufficiently rigorous. Essential details such as dataset preparation, preprocessing steps, modifications to the YOLOv8 architecture, training procedures, hyperparameter choices, and evaluation metrics are either missing or superficially described. Mathematical formulations, where present, are poorly formatted, and equations lack a proper definition of variables. A step-by-step description of the entire pipeline, from image acquisition to model evaluation, as well as correctly typeset equations and clearly defined terms, would substantially improve this section.]

Response 3:[

We deeply appreciate your incisive critique of the manuscript's limitations—your observations are entirely valid. Regarding methodological details:

• Hyperparameter specifications for the core model (DW-YOLO) are explicitly documented in lines 335–343, including learning rate (0.01), batch size (16), and anchor optimization.
• All other parameters retain YOLOv8's default configurations (e.g., SGD momentum=0.937, weight decay=0.0005), as validated against the original implementation per Ultralytics YOLOv8 benchmarks.
• We acknowledge that earlier methodological descriptions lacked sufficient technical granularity Your guidance has significantly strengthened the scholarly robustness of this work.

]

Comments 4:[Another limitation is the lack of robust experimental validation. The performance evaluation and ablation studies do not include meaningful comparisons with baseline models or state-of-the-art methods. Without such benchmarks, it is difficult to determine whether the proposed modifications to YOLOv8 offer real improvements. A comparative analysis with standard approaches would strengthen the claims.]

Response 4:[We acknowledge your valid point that datasets should ideally be made publicly available to advance research. However, due to industrial application considerations in this study, immediate public release of the dataset is not feasible. We will make both code and dataset publicly accessible after the completion of this research project. Regarding the sample size, annotation protocols, and data partitioning you mentioned, these aspects involve another researcher's data processing work. Since their related paper has not yet been published, we cannot disclose extensive details. We can only release data within the scope of our personally responsible research components. Thank you for your understanding.]

Comments 5:[The description of the experimental apparatus is also insufficient. Although Figure 7 illustrates the setup, there is no detailed account of its design or specifications. Important parameters such as lighting conditions, camera resolution, distance from the sample, and calibration procedures are omitted, yet these factors directly affect image quality and, consequently, model performance. Providing comprehensive specifications, diagrams, and a clear operational workflow would be essential for reproducibility.]

Response 5:[We have revised the experimental setup description in lines 312–318. Due to the ongoing patent application for this apparatus, we are unable to disclose further technical details as it involves intellectual property rights of our laboratory colleagues. Once the patent authorization is granted in the near future, we will be permitted to reveal additional specifications of the apparatus. We kindly request your understanding regarding this constraint]

Comments 6:[In addition, the manuscript does not indicate whether the dataset or source code will be made publicly available. Transparency and reproducibility are essential in modern scientific research and sharing both would allow others to build upon this work. Details about the dataset, such as number of samples, resolution, labeling protocol, class distribution, and data splits, should be included, along with links to public repositories hosting the data and source code.]

Response 6:[First, I acknowledge the issue you pointed out. Theoretically, all datasets should be made public so that others can further advance research based on them. However, since this study involves upcoming industrial applications, it is not convenient to make the dataset publicly available at this time. After the completion of this research project, I will make both the code and dataset public. Regarding the sample size, annotation protocols, and data partitioning you mentioned, these involve the data processing work of another researcher in this experiment. Since his paper has not yet been published, I cannot disclose too much information. I can only make public the data within the scope of my responsible research area. Thank you for your understanding.]

Comments 7:[Finally, the overall presentation of the figures requires improvement. Several images are low-quality, cropped, blurry, or contain incomplete and unreadable text. This undermines the clarity of the evidence presented. All figures should be revised to meet high-resolution standards and comply with the journal’s formatting requirements.]

Response 7:[Thank you for your detailed review of the image-related content in this manuscript. I have reprocessed all images for enhanced clarity and re-uploaded them. I kindly request your review of these revised visuals.]

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

Please follow the following points carefully.

• Introduction section is too confusing, there are set of different objectives in each paragraph that leads to unsure which one to follow. For example, (1) we propose an identification system integrating, (2) enhances production efficiency, accuracy, and timeliness. (3) enhancing the effectiveness of computer vision detection tasks. (4) to boost detection performance, reducing inference time and memory usage in object detection algorithms. (5) we propose an improved DW-YOLO strategy to address the detection challenges (6) improving overall accuracy and enabling more effective. Please be specific in stating the main objectives.

• Please unify the word figure within the manuscript, Figure 9 and 10 are presented as “Fig” while others are written “Figure”
• Tables and figures and not formatted well.
• Conclusion should summarized the main finding (from tables and figures) in numerical presentation.

Author Response

Comments 1:[Introduction section is too confusing, there are set of different objectives in each paragraph that leads to unsure which one to follow. For example, (1) we propose an identification system integrating, (2) enhances production efficiency, accuracy, and timeliness. (3) enhancing the effectiveness of computer vision detection tasks. (4) to boost detection performance, reducing inference time and memory usage in object detection algorithms. (5) we propose an improved DW-YOLO strategy to address the detection challenges (6) improving overall accuracy and enabling more effective. Please be specific in stating the main objectives.]

Response 1:[We sincerely appreciate your professional insights and fully acknowledge their value. The introduction of this paper adheres to the following logical framework: We propose an enhanced model that serves as the core component of the recognition system. The improvements in both model performance and inference speed are specifically targeted at strengthening the feasibility of industrial applications in the graphite sector, thereby establishing a technical foundation for production deployment of graphite ore grade identification tasks. We have strategically supplemented comparative analysis of key literature in lines 105-115 and added a dedicated "Related Work" section spanning lines 150-154.]

Comments 2:[Please unify the word figure within the manuscript, Figure 9 and 10 are presented as “Fig” while others are written “Figure”]

Response 2:[We sincerely appreciate your professional insights and fully acknowledge their value. A modification has been made on line 367 and 371.]

Comments 3:[Tables and figures and not formatted well.]

Response 3:[We appreciate your valuable feedback and have duly revised the formatting of all tables and numerical values therein to meet academic standards.]

Comments 4:[Conclusion should summarized the main finding (from tables and figures) in numerical presentation.]

Response 4:[Thank you for your insightful critique—we have implemented the revisions in lines 458–472 to unify the introduction's objective framework.]

 

Author Response File: Author Response.docx

Reviewer 5 Report

Comments and Suggestions for Authors

This manuscript proposes a YOLO-based model to identify the ore grade of different graphene ores. The followings questions and comments should be addressed during revision:

1. What does DW stand for in DW-YOLO?
2. The math font and equations formatting need some attention, they are not properly formatted in this submitted version.
3. Line 299-300, are there any data augmentation techniques applied to account for different light conditions?
4. How sensitive is the proposed model to the orientation of the ore when imaged? If the same ore is imaged multiple times at multiple different orientations, will the model provide similar/identical classification?
5. Figure 9 is outside of the bounds of the page width.
6. Table 3-3, the model training should be repeated multiple times to obtain mean and standard deviation of the performance metrics.
7. Figs 11 and 12 also need to be re-formatted.

Author Response

评论1:[DW-YOLO的DW代表什么？]

Response 1:[DW stands for "Deep Work," referring to a more in-depth working approach with the YOLO model.]

Comments 2:[The math font and equations formatting need some attention, they are not properly formatted in this submitted version.]

Response 2:[I have made format revisions on lines 246, 271, 287, and 292 of the article.]

Comments 3:[Line 299-300, are there any data augmentation techniques applied to account for different light conditions?]

Response 3:[Data augmentation techniques were not employed to account for varying lighting conditions. The complete workflow of this project also involves the work of other laboratory personnel, who apply deblurring algorithms to the images to ensure that they meet the experimental requirements for detection.]

Comments 4:[How sensitive is the proposed model to the orientation of the ore when imaged? If the same ore is imaged multiple times at multiple different orientations, will the model provide similar/identical classification?]

Response 4:[In the dataset created for this paper, each experimental sample was photographed from multiple angles, and the resulting classifications exhibited similarity. Below are examples of images captured from different directions.]

Comments 5:[Figure 9 is outside of the bounds of the page width.]

Response 5:[A modification has been made on line 369.]

Comments 6:[Table 3-3, the model training should be repeated multiple times to obtain mean and standard deviation of the performance metrics.]

Response 6 :[A modification has been made on line 411.]

Comments 7:[Figs 11 and 12 also need to be re-formatted.]

Response 7:[A modification has been made on line 413 and 428.]

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

There is no further comment. I recommend that this paper is published.

Author Response

We have great respect for your position as the gatekeeper of scientific quality, and thank you again for your insightful comments and consideration of our revisions.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have made significant improvements in this work. Some technical terms and misconceptions have been clarified. Plots and figures are now better. Changing from graphene to graphite makes the manuscript clearer. Many of the methodological observations were corrected. Furthermore, the concern about the clarity of the purpose of the DW-YOLO algorithm for identifying grades of graphite ore has been addressed.

However, many comments were not adequately revised:

1. In my previous review, I recommended including a “related work” section. The authors included some new references and edited lines 105-115. But, in my opinion, this is insufficient as a framework to compare the work with. It is strongly recommended to critically evaluate the previous efforts to reveal the research gap that the authors want to cover. It includes not only qualitative descriptions but also quantitative data. For example, the authors argue that the work of Xiang et al. requires high-quality imagery and ignores multi-perspective orientations. However, what is the problem behind that? Does it affect the graphite ore grading? If yes, what is the impact of that, and how is your approach better to solve this issue? The establishment of the state-of-the-art (related work) and the further comparison with the results are essential elements to demonstrate the progress of science and technology. Thus, it is not clear to me how this work contributes to the main problem (graphite ore grade identification) and how it builds upon previous efforts.
2. In the methodology section, the manuscript still presents unformatted equations with undeclared variables. It is critical that a well-formulated mathematical background be provided to understand the methods used properly. This observation was ignored.
3. My comment 4 (about the lack of robust experimental validation and comparison with benchmark methods) was also ignored. In their response, the authors do not provide details of the justification for it. I still recommend evaluating the quantitative results and comparing them against others. It is related to the first comment of this review also.
4. Regarding the experimental apparatus, although I understand that not all details can be provided due to the patent process, I think that Figure 7, which only depicts a box with lamps, cables, and fans, is not helpful. Then, I recommend deleting it if you are not willing to explain how this apparatus contributes to your results.
5. I understand that other researchers are responsible for data acquisition and dataset curation. However, it is very necessary to include at least a general but detailed enough description of the dataset: number of classes, number of samples per class (to analyze if it is a balanced dataset), examples of each class (Fig. 8 only exhibits 3 samples, and it is not clear what surface characteristics belong to each graphite grade). Also, it is unclear whether any pre-processing steps are performed, such as standardizing image resolution or other common operations.

Please consider that providing an adequate description of the dataset is essential for understanding the contribution of the algorithms and their functionality. Failing to provide a full context does not help to position your work within the state of the art. You, the authors, acknowledge that “… we will expand the image datasets to cover more mining sites and further promote the intelligent and automated recognition of ore grades.” So you understand the importance of the dataset for this kind of work and ensuring the reproducibility of science.

Additionally, the authors declare that “… the surface features of the ore cannot directly indicate its carbon content. Therefore, after image capture, a carbon-sulfur analysis process and graphite carbonization measurement workflow were performed.” However, it is unclear how this information is introduced to the workflow for the final determination of graphite ore grade.

Moreover, I recommend carefully revising the equations to explain how TP, FP, and so on were interpreted, considering that you have six grades (classes?) of graphite ore.

Thus, I still consider that “The methodology section is incomplete and insufficiently rigorous”, as mentioned in my previous review.

6. Regarding my comments 4, 5, and 6 from my previous review, the authors essentially state that they cannot provide more information due to patent restrictions, industrial processes involved, and ongoing research. Again, I want to remind you that reproducibility and transparency are essential for scientific progress. So, if you cannot make an effort to provide enough information to understand and reproduce your research without compromising the confidentiality of these processes, maybe you should consider waiting until those processes are complete before publishing your research. Please consider this comment with all my respect and devotion to the authors and their work, which I find interesting and helpful if sufficient provisions are made to complete this manuscript.

Author Response

Comments 1:[In my previous review, I recommended including a “related work” section. The authors included some new references and edited lines 105-115. But, in my opinion, this is insufficient as a framework to compare the work with. It is strongly recommended to critically evaluate the previous efforts to reveal the research gap that the authors want to cover. It includes not only qualitative descriptions but also quantitative data. For example, the authors argue that the work of Xiang et al. requires high-quality imagery and ignores multi-perspective orientations. However, what is the problem behind that? Does it affect the graphite ore grading? If yes, what is the impact of that, and how is your approach better to solve this issue? The establishment of the state-of-the-art (related work) and the further comparison with the results are essential elements to demonstrate the progress of science and technology. Thus, it is not clear to me how this work contributes to the main problem (graphite ore grade identification) and how it builds upon previous efforts.]

Response 1:[Thank you for your suggestions on this article, I think your guiding advice on this article is very pertinent and professional. I modified lines 106-150.]

Comments 2:[In the methodology section, the manuscript still presents unformatted equations with undeclared variables. It is critical that a well-formulated mathematical background be provided to understand the methods used properly. This observation was ignored. ]

Response 2:[Thank you for your suggestions on this article, I think your guiding advice on this article is very pertinent and professional. I modified lines 247and273-276.]

Comments 3:[My comment 4 (about the lack of robust experimental validation and comparison with benchmark methods) was also ignored. In their response, the authors do not provide details of the justification for it. I still recommend evaluating the quantitative results and comparing them against others. It is related to the first comment of this review also.]

Response 3:[Thank you for your suggestions on this article, I think your guiding advice on this article is very pertinent and professional. I modified lines 397-422.]

Comments 4:[Regarding the experimental apparatus, although I understand that not all details can be provided due to the patent process, I think that Figure 7, which only depicts a box with lamps, cables, and fans, is not helpful. Then, I recommend deleting it if you are not willing to explain how this apparatus contributes to your results. ]

Response 4:[Thank you for your suggestions on this article, I think your guiding advice on this article is very pertinent and professional. I modified lines 106-150.]

 

Comments 5:[I understand that other researchers are responsible for data acquisition and dataset curation. However, it is very necessary to include at least a general but detailed enough description of the dataset: number of classes, number of samples per class (to analyze if it is a balanced dataset), examples of each class (Fig. 8 only exhibits 3 samples, and it is not clear what surface characteristics belong to each graphite grade). Also, it is unclear whether any pre-processing steps are performed, such as standardizing image resolution or other common operations. ]

Response 5:[The data collection and standardization processes were both completed using this specific equipment. Although additional technical details cannot be disclosed due to proprietary constraints, this apparatus represents an essential component of the entire workflow. I therefore recommend retaining this figure in the manuscript, as it significantly enhances the completeness and methodological transparency of our research. The inclusion of Figure 7 provides critical context for understanding our data acquisition pipeline, which is fundamental to the reproducibility and validity of our experimental results.]

 

Comments 6:[Regarding my comments 4, 5, and 6 from my previous review, the authors essentially state that they cannot provide more information due to patent restrictions, industrial processes involved, and ongoing research. Again, I want to remind you that reproducibility and transparency are essential for scientific progress. So, if you cannot make an effort to provide enough information to understand and reproduce your research without compromising the confidentiality of these processes, maybe you should consider waiting until those processes are complete before publishing your research. Please consider this comment with all my respect and devotion to the authors and their work, which I find interesting and helpful if sufficient provisions are made to complete this manuscript.]

Response 6:[Thank you for your thoughtful and constructive feedback regarding our manuscript. We sincerely appreciate your dedication to scientific rigor and your commitment to maintaining the highest standards of reproducibility and transparency in academic research. Your comments have significantly strengthened our work, and we deeply value your expertise and the time you've invested in reviewing our paper.Response to Concerns About Information Disclosure

We fully acknowledge and respect your emphasis on reproducibility as a cornerstone of scientific progress. Please accept our assurance that we have made every possible effort to provide sufficient methodological details while navigating the constraints imposed by:

1. Patent protection requirements for novel industrial processes currently under review
2. Confidentiality agreements with industry partners regarding proprietary mining and processing techniques
3. Ongoing research collaborations where full disclosure could compromise future studies

We deeply respect your position as a gatekeeper of scientific quality, and we are confident that the enhanced manuscript now provides sufficient transparency to support its scientific contribution while honoring our commitments to industry partners.

Thank you again for your insightful comments and for considering our revisions. We would be grateful for the opportunity to have our substantially improved work reevaluated.]

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

Dear Editor and Authors, 

The authors have addressed all the required comments. The manuscript looks ready for acceptance. 

Regards, 

Ali

Author Response

We have great respect for your position as the gatekeeper of scientific quality, and thank you again for your insightful comments and consideration of our revisions.

Round 3

Reviewer 3 Report

Comments and Suggestions for Authors

I have no more comments. Thanks. 

I recommend revising the formatting of equations and the quality of tables and figures. 

Author Response

comments 1:[I recommend revising the formatting of equations and the quality of tables and figures. ]

Response 1:[Thank you for your valuable review comments. I have reformatted the tables and uploaded higher-quality images accordingly. We sincerely appreciate your thorough and thoughtful feedback, which has greatly contributed to improving the clarity and overall quality of our manuscript. Your expertise and dedication are truly valued.]

 

Author Response File: Author Response.docx