Intelligent Classifier for Identifying and Managing Sheep and Goat Faces Using Deep Learning

Response 2: Agree. We have, accordingly, included example of face images used in the training, in a new figure added to the article. (Method section para 2 line 165-170 Fig 8 and Fig 9)

Comments 3: The title mentions about the managing bovine based on face but there is no discussion on how the animals are being managed through face identification.

Response 3: Understood. Considering your suggestion discussion has been included in the manuscript (Please see paragraph 7 and 8 of Introduction section, line 59-69 and 74-78)

Comments 4: The introduction is very lengthy and review of literature is included in it.

Response 4: Thank you for the review. The introduction section was reworked to address this demand. The review of literature was split from the introduction. (Please see sub section 1.1 Literature Review line number 79-159)

Comments 5: The last paragraph of the introduction claims so many benefits but did not explain any of these how the face identification can actually help the farmers in these areas and how it will be executed.

Response 5: Agreed. The cited paragraph was rewritten to address the explanation of the mentioned benefits.

Comments 6: Page-5, parameters TN/FN are not clearly defined and not explained in reference to understanding of a layman.

Response 6: Thank you. Indeed, the parameters were not clearly defined neither explained. It was corrected. (Please see the paragraph 8 of Method section line number 191-192)

Comments 7: All images shown in the paper are not of good quality.

Response 7: We understand. The quality of all images were improved and some figures were reworked to provide better visualization.

Comments 8: The source of data is not clearly mentioned i.e. from where images have been taken whether these are original pictures taken by the researchers or taken from third party.

Response 8: We are sorry for this mistake. Here we have used various goat and sheep datasets available on the internet and Kaggle datasets. In our dataset, we have tried to include a variety of images of goats and sheep of all possible colors and breeds. (Method section para 2 line 165-170 Fig 8 and Fig 9)

Comments 9: Result of evaluation parameters has not been explained from the view point of farmers whereas the paper is focused on the benefits for the farmers.

Response 9: Indeed. We reworked the results sections to address this demand.

Comments 10: Conclusion is poorly written.

Response 10: Indeed. The conclusions were rewritten completely. We hope this new version is more suitable.

Comments 11: The authors have claimed (2-3 times in the paper) that the results of this work will improve the comfort of livestock but have not explained how?

Response 11: Thank you for your consideration. We addressed this demand by explaining with more details the benefits of the results and its possible applications.

Comments 12: The claims made in the last para of the conclusion does not have any relevance with the work presented in the paper.

Response 12: Understood. As already mentioned, the conclusions were rewritten in order to better explain the benefits. Any claims without relevance were removed.

Comments 13: The authors have used a number of standard algorithms of deep learning to test the results but have not explained the basis of selection of so many algorithms.

Response 13: That is true. We understand that many algorithms were presented, but our initial idea was to compare as many techniques as possible, thus the explanation.

Comments 14: Could not find any novelty related to the data set creation of face images. 

Response 14: We used a public dataset. (Method section para 2 line 165-170 Fig 8 and Fig 9)

4. Response to Comments on the Quality of English Language

Point 1: Need improvement in English language of the paper

Response 1:    The text has been completely revised and grammatical and spelling errors were corrected.

5. Additional clarifications

All of the reviewers demands were addressed.

Comments 1: The initial dataset is relatively small, requiring considerable effort to optimize the model. The artificial data augmentation techniques may not fully capture real-world variations, and the application of the model to other animal species is limited.

Response 1: Using artificial data augmentation techniques is a common practice in data science and has been shown to significantly improve the performance of machine learning models.  (Please paragraph 3 of subsection 3.1 Roboflow YOLO line number 344-352)

Comments 2: Additionally, the paper lacks detailed information on the computational resources and time required, making it difficult to understand the practical costs of the study.

Response 2: Agree. Regarding this we included Table 10.

Comments 3: Furthermore, despite the existence of numerous studies on facial recognition in livestock, the paper does not adequately justify the need for this specific study or explain why it is important to distinguish between goats and sheep.

Response 3: Understood. Thank you for the review. The introduction section was reworked to address this demand.

Comments 4: The findings are not thoroughly discussed or compared with existing literature.

Response 4: Thank you for the review. Regarding this we included Table No. 1

Comments 5: Moreover, the literature review section is not well-formatted and only a limited number of studies are reviewed, which weakens the foundation of the research.

Response 5: Agree. Thank you for the review. The introduction section was reworked to address this demand. The review of literature was split from the introduction.

4. Response to Comments on the Quality of English Language

Point 1: English language fine. No issues detected.

5. Additional clarifications

All the reviewer’s demands were addressed.

Comments 1: Page 4: The authors mentioned that they increased the total number of images to 35,204 using a data augmentation tool, but they did not provide details on the specific methods or techniques used for data augmentation. It is crucial to describe the data augmentation process, including the types of transformations applied and any parameters or settings used.

Response 1: Thank you for pointing this out. I agree with this comment. Details are included in the manuscript. (Please paragraph 3 of subsection 3.1 Roboflow YOLO line number 344-352)

Comments 2: Page 4: Please provide a reference for the recall and precision formula. You can refer to the following publication: https://doi.org/10.1016/j.jspr.2021.101800

Response 2: Agree. Thank you. We have included the formulas. (Please see the paragraph 8 of Method section line number 191-192)

Comments 3: Page 6: The hardware and software configuration of this research should be described.

Response 3: Understood. We described the hardware and software in the manuscript. (Please see the last paragraph of Method Section Line number 226-232)

Comments 4: Results section: The superiority of one model over another should be justified, taking into account the structure and topologies of the DL models.

Response 4: Thank you for the review. Comparison of models have been described in the result section.

Comments 5: It is necessary to compare the obtained results with similar articles.

Response 5: Agree.  Comparisons of the models have been done in the result section.

4. Response to Comments on the Quality of English Language

Point 1: English language fine. No issues detected.

5. Additional clarifications

All of the reviewers demands were addressed.