Chicken Tracking and Individual Bird Activity Monitoring Using the BoT-SORT Algorithm

Round 1

Reviewer 1 Report

1. The introduction is not detailed enough and an objective analysis of the current state of research, and a comparison of similar studies in the field.

2. In the second subsection, it is important to state when and where the data was collected. In addition, challenges encountered in data detection, such as masking or other related difficulties, should be accounted for.

3. Problem with paragraph Indent in paragraph of subsection 3.4 in the text. Rather, authors should carefully review and correct similar errors throughout the official document.

4. The description of Figure 3 lacks sufficient detail and should be expanded to provide a comprehensive explanation of the role of the curve image in the context of the study.

5. To further demonstrate the performance of the BoT-SORT algorithm under study, it is recommended to include additional relevant tested object tracking algorithms as a comparative analysis. This will provide a comprehensive assessment of the effectiveness of the BoT-SORT algorithm compared to other algorithms.

Author Response

We appreciate the reviewer's time and dedication in evaluating our article. Below we answer all the questions asked point by point. We remain at your disposal for other corrections, if necessary.

1. The introduction is not detailed enough and an objective analysis of the current state of research, and a comparison of similar studies in the field.

We have expanded the Introduction chapter.

2. In the second subsection, it is important to state when and where the data was collected. In addition, challenges encountered in data detection, such as masking or other related difficulties, should be accounted for.

We added more information about the experiment. However, it is possible to obtain more details from the original article, published by Fernandes et al. (2021).

3. Problem with paragraph Indent in paragraph of subsection 3.4 in the text. Rather, authors should carefully review and correct similar errors throughout the official document.

Done.

4. The description of Figure 3 lacks sufficient detail and should be expanded to provide a comprehensive explanation of the role of the curve image in the context of the study.

Done.

5. To further demonstrate the performance of the BoT-SORT algorithm under study, it is recommended to include additional relevant tested object tracking algorithms as a comparative analysis. This will provide a comprehensive assessment of the effectiveness of the BoT-SORT algorithm compared to other algorithms.

We agree with the reviewer, but we understand that our dataset is not broad and diverse enough to promote an adequate comparison between different tracking models. In this article we propose to present the use of this new tracking model, with a high success rate, in a proof of concept. We understand that comparing efficiency between models depends on a more robust dataset.

Reviewer 2 Report

I am expressing my gratitude for the opportunity to review the article titled "Tracking chickens in flocks and estimating individual distance traveled using the BoT-SORT algorithm." However, I also have some critical points and concerns about the paper:

General suggestions:

1) Please be consistent regarding the aim of the study (lines 54 - 56; line 243).

2) The introduction is very short and more detailed information needs to be added.

3) Materials and Methods: Please include relevant informations regarding the housing conditions (nest, feeding facilities,...) as well as age of layers and period of investigation in relation to the light program.

4) Why did you track the behaviour of individuals as the group behaviour is important for monitoring the welfare of layers under commercial conditions (line 365)?

5) Discussion: The discussion needs fundamental improvements as your results. It is kind of an extended introduction. Your results must be set into relation to the results of other studies in this field.

6) Results: Number of occurence and times spend by birds running, exploring and resting are shown in table 2. How are these values to be interpreted? Is there a corridor of expected values indicating a good status quo of animal welfare?

Detailed suggestions:

1) line 293: Please explain, why exactly a minimum of 18 frames are required?

2) Layers are often kept in aviaries. Which improvments are required with the aim of implementing the descripted algorithm in welfare monitoring in commercial production systems? 

3) Is there an impact on the accuracy of tracking of hens regarding the type of bedding and plumage (colour, optical contrast)?

4) Is there an impact on the accuracy of tracking of hens regarding the tlight intensity and light spectrum (lighting of the stable)?

Based on these concerns, I recommend publishing the manuscript after minor revisions for publication.

Author Response

I thank the reviewer for his willingness to evaluate our article and propose corrections that improve the understanding of our study. In the following, I hope to have clarified all doubts and remain at your disposal for further corrections, if necessary.

General suggestions:

1) Please be consistent regarding the aim of the study (lines 54 - 56; line 243).

We rewrote the objectives of this study.

2) The introduction is very short and more detailed information needs to be added.

We added new studies to the introduction, expanding it considerably.

3) Materials and Methods: Please include relevant informations regarding the housing conditions (nest, feeding facilities,...) as well as age of layers and period of investigation in relation to the light program.

We added more information about the experiment. However, it is possible to obtain more details from the original article, published by Fernandes et al. (2021).

4) Why did you track the behaviour of individuals as the group behaviour is important for monitoring the welfare of layers under commercial conditions (line 365)?

We identify individual chicken behaviors because this becomes challenging when birds are grouped together. This is due to the limitations of computer vision systems in separating and classifying birds in low-resolution images. Our system has the ability to track each bird for long periods of time in the video, as long as the chicken remains visible in the recording.

5) Discussion: The discussion needs fundamental improvements as your results. It is kind of an extended introduction. Your results must be set into relation to the results of other studies in this field.

We expanded the discussion.

6) Results: Number of occurence and times spend by birds running, exploring and resting are shown in table 2. How are these values to be interpreted? Is there a corridor of expected values indicating a good status quo of animal welfare?

We found no reference values in the literature. Our hypothesis is that it will be possible to create these reference values from the precise extraction of measurements of distance and movement speed of chickens. We added explanatory text to the article.

Detailed suggestions:

1) line 293: Please explain, why exactly a minimum of 18 frames are required?

We have added the following explanatory text.

“The limit of 18 frames was chosen based on observations since birds frequently performed movements that caused abrupt changes in the image centroid.”

2) Layers are often kept in aviaries. Which improvments are required with the aim of implementing the descripted algorithm in welfare monitoring in commercial production systems?

We sought to develop a system that would be applicable under commercial conditions. As described in the manuscript, we chose to consider a bird that leaves the scene and returns as a new bird, as we imagine that under commercial conditions it will not be possible to monitor the entire aviary and "blind spots" are inevitable.

3) Is there an impact on the accuracy of tracking of hens regarding the type of bedding and plumage (colour, optical contrast)?

Yes. But for situations different from those addressed in our work, a new YOLO model for detection must be developed.

4) Is there an impact on the accuracy of tracking of hens regarding the tlight intensity and light spectrum (lighting of the stable)?

Yes, but the previous answer applies in this case too. Situations in which there are variations in intensity and/or light spectrum must be included in the dataset for training and validating the detection model.

Reviewer 3 Report

1.     The article titled “Tracking chickens in flocks and estimating individual distance travelled using the BoT-SORT algorithm” presented an interesting study that involved the application of AI, deep learning and Machine Learning/ machine vision for monitoring poultry birds in the flocks. This study presented a new algorithm and model to identify the individual birds and track and measure their activity to monitor them for their welfare, producing good convincing results with better efficiency. Therefore, the study has scope and value for the relevant individuals and communities.

2.     However, the quality of the presentation is low and must be improved for consideration in Journal publication. All sections of the article need thorough rigorous revision and editing for better improvement according to Journal standard requirements.

3.     Title: The title needs to be revised to make it short, comprehensive and catchy.

4.     Suggestion: Chicken tracking and individual bird activity monitoring in flocks using the BoT-SORT algorithm.

5.     Abstract: The abstract should be in a systematic structured format (background/ need, intervention/ methods and technologies, and key outcomes/ important results).

6.     Introduction: The introduction is too much short, properly improve this section with some details of the study background, research problems/ questions and study needs, supporting and relating with existing literature/citations. Appropriate introduction/ description of methods, tools and technologies that were involved in the study. Main purpose and objectives of the study.

7.     Materials and Methods: Some corrections/ improvements are required.

8.     Line 67: The images have a resolution of 352x240 pixels and the capture rate was 30 fps. Is 352x240 pixels the size of the image or the resolution of the image? How to define and select this resolution? 352x240=84480 pixels is low resolution and may not work well for tracking.

9.     May describe the boundary conditions and limitations of applied models.

10. Results: The quality of the presentation and write-up of the section needs to improve.

11. Figure 9. Separate the image groups and clearly denote with letters a and b below in the centre.

12. Conclusion: It is short and only describes the general outcomes of the study. This is not the pattern of conclusion. The conclusion should be a well-organized and standalone summary of the whole research, starting from a wider background, describing essential aspects of the study and ending with concise outcomes and a way forward as future directions.

Overall write-up of the article needs to improve rigorously, using inclusive language with appropriate choice of words, and well-structured sentences and paragraphs.

I recommended revising, improving the article thoroughly, and submitting it after revision so that it would be considered for publication.

Overall write-up of the article needs to improve rigorously, using inclusive language with appropriate choice of words, and well-structured sentences and paragraphs.

Author Response

1. The article titled “Tracking chickens in flocks and estimating individual distance travelled using the BoT-SORT algorithm” presented an interesting study that involved the application of AI, deep learning and Machine Learning/ machine vision for monitoring poultry birds in the flocks. This study presented a new algorithm and model to identify the individual birds and track and measure their activity to monitor them for their welfare, producing good convincing results with better efficiency. Therefore, the study has scope and value for the relevant individuals and communities.
2. However, the quality of the presentation is low and must be improved for consideration in Journal publication. All sections of the article need thorough rigorous revision and editing for better improvement according to Journal standard requirements.

We appreciate the reviewer's time in evaluating our manuscript. Below, we answer the questions point by point and provide the improvements requested in the main text.

3. Title: The title needs to be revised to make it short, comprehensive and catchy.
4. Suggestion: Chicken tracking and individual bird activity monitoring in flocks using the BoT-SORT algorithm.

We followed the suggestion and changed the title to make it shorter and more attractive.

5. Abstract: The abstract should be in a systematic structured format (background/ need, intervention/ methods and technologies, and key outcomes/ important results).

We rewrote the abstract.

6. Introduction: The introduction is too much short, properly improve this section with some details of the study background, research problems/ questions and study needs, supporting and relating with existing literature/citations. Appropriate introduction/ description of methods, tools and technologies that were involved in the study. Main purpose and objectives of the study.

We have expanded the Introduction chapter, with the aim of presenting similar studies for comparison.

7. Materials and Methods: Some corrections/ improvements are required.
8. Line 67: The images have a resolution of 352x240 pixels and the capture rate was 30 fps. Is 352x240 pixels the size of the image or the resolution of the image? How to define and select this resolution? 352x240=84480 pixels is low resolution and may not work well for tracking.

We answer the question in the text. The dimension 352x240 pixels refers to the size of the image.

9. May describe the boundary conditions and limitations of applied models.

We consider it more appropriate to present all the limitations of our study, including the limitations of the model, in chapter 4.1.

10. Results: The quality of the presentation and write-up of the section needs to improve.
11. Figure 9. Separate the image groups and clearly denote with letters a and b below in the centre.

Done.

12. Conclusion: It is short and only describes the general outcomes of the study. This is not the pattern of conclusion. The conclusion should be a well-organized and standalone summary of the whole research, starting from a wider background, describing essential aspects of the study and ending with concise outcomes and a way forward as future directions.

We completed the conclusions as suggested by the reviewer.

Round 2

Reviewer 3 Report

I appreciate the improvements made, but it is emphasised that reconsider the conclusion part to revise, it is not structured well and has not improved up to the mark. It must be improved into a well-structured form by supporting it with the results. The author guide of the journal and good published paper could provide clear ideas to improve.

Minor grammar corrections are required.

Author Response

We thank the reviewer once again for your attention. We rewrote the conclusions, supporting them with the results obtained. I remain at your disposal for additional corrections, if necessary.